How artificial

intelligence can
power your
climate action
strategy
 Executive summary
As the COVID-19 pandemic spread across the world, II. AI-enabled use cases are already reducing
it also highlighted another major collective hazard GHG emissions and can accelerate climate
we face – climate change. Extreme weather events action
are increasing, lives are being affected and placed in
danger, and costs are mounting for governments and • Across sectors, AI-enabled use cases have helped
industry. Given the urgency of this issue, this report organizations reduce GHG emissions by 13% and
seeks to understand how AI can accelerate improve power efficiency by 11% in the last two
our response. years. AI use cases have also helped reduce waste
and deadweight assets by improving their utilization
Our research found that: by 12%.
I. AI offers many climate action use cases • Our modelling estimates that, by 2030, AI-enabled
use cases have the potential to help organizations
• We analyzed over 70 AI-enabled use cases for
fulfil 11–45% of the ‘Economic Emission Intensity’
climate action and identified the ten most impactful
targets of the Paris Agreement, depending on the
ones. By that we mean they offer significant
scale of AI adoption across sectors. For instance, for
benefits for organizations in terms of reduced
the automotive sector, AI-enabled use cases have
greenhouse gas (GHG) emissions, improved energy
the potential to deliver 8 percentage points of the
efficiency, and reduced waste.
37% reduction (more than one-fifth) required
Examples include:
by 2030.
– Tracking GHG emissions and tracing GHG leakages
• In the future, AI is expected to reduce GHG
at industrial sites
emissions by 16% and improve power efficiency by
– Improving the energy efficiency of facilities and 15% in the next three to five years.
industrial processes
III. Even though climate action is a strategic
– AI for designing new products that reduce waste priority, most organizations are struggling to
and emissions during prototyping, production, support climate action with AI capabilities
and usage
• 67% have made long-term business decisions to
– AI for inventory management - improving tackle the consequences of climate change.
demand planning and reducing wastage of food
products and raw materials • However, few organizations are successfully
combining their climate vision with AI capabilities
– Route optimization and fleet management for and have driven solutions to scale. We call these
retail, automotive, and consumer products firms high-performing organizations – Climate AI

2 AI in Climate Change
 Champions. They represent only 13% of the entire • Collaborate with the climate action ecosystem.
survey sample.
• Harness AI to bring greater focus in reducing scope
• A closer look at the challenges that organizations 3 emissions.
face in scaling AI for climate action reveals a range
The research approach draws on a range of tactics,
of issues: only 42% of the potential AI use cases
from using climate models developed in partnership
are being experimented with and more than eight
with 'right. based on science' – a climate startup –
in ten organizations spend less than 5% of climate
for estimating the impact of AI on greenhouse gas
change investment on AI.
(GHG) emissions to a survey of over 400 sustainability
IV. How organizations can leverage AI’s full executives and 400 business/tech executives from
climate action potential the automotive, industrial/process manufacturing,
energy & utilities, and consumer products and
We believe that six action areas are critical: retail industries. Additionally, we surveyed 300
• Account for, and take measures to combat, the experts, including regulators, academics, and non-
negative impact of AI on the climate. governmental organizations working in the field
of sustainability, and interviewed over 40 industry
• Educate sustainability teams on how AI can make executives and experts. More details on the research
a real difference and educate AI teams on the methodology are in the Appendix.
criticality of climate change.
• Lay down the technological foundations for
AI-powered climate change action.
• Scale use cases on the basis of impact for your
sector and emissions intensity of particular
functions.

3
Introduction
The climate crisis is urgent and the organizations need In recent years, we have witnessed a range of unprecedented
to act now natural disasters as a result of major changes in the earth’s
climate. A UN report released in October 2020 states that
Today’s global pandemic is not the only global threat we face
between 2000–2019, there were 7,348 major recorded
that has wide-ranging impact on economies and society. The
disaster events claiming 1.23 million lives, affecting 4.2 billion
extent of climate crisis is laid bare by the changes we are
people resulting in approximately USD2.97 trillion in global
seeing:
economic losses.6
• The atmospheric levels of CO2 are at the highest compared
Large organizations have a sizeable climate footprint and
to any point in human history (see Figure 1).1 This is
their operations are adding to GHG concentration, which
despite the steep drop in global GHG emissions during the
is the prime reason behind global warming and climate
pandemic lockdown.2
change. If we look at four sectors that make up our research –
• Average temperatures have been on the rise for decades automotive manufacturing, industrial manufacturing, energy
now.3 & utilities, and consumer products and retail – they together
contribute roughly 50% of the world’s total GHG emissions
• Average sea levels have risen about 8 inches (over 20 (see Figure 1).
centimeters) since 1880 and 3 inches in the last 25 years
alone.4
• 2019 was also the hottest year on record for the world’s
oceans.5

4 AI in Climate Change
Figure 1. Rising CO2 levels and sector distribution of GHG emissions

Carbon dioxide in atmosphere at record level

Mauna Loa Observatory measured a record monthly average atmospheric carbon dioxide concentration in May,
typically the peak of the year.

417.1 parts per million

Trend (adjusted for

seasonal patterns)
410 parts per million

Detail
400

400

200

390
0
2010 2015 2020 1960 2020

SOURCE: NOAA Global Monitoring Laboratory JOHN MUYSKENS/THE WASHINGTION POST

Source: The Washington Post, “Earth’s carbon dioxide levels hit record high, despite coronavirus-related emissions drop,” June 2020.

Breakdown of global greenhouse gas emissions – 2016

Consumer products & retail

Road/transport

8% 12%
Discrete
Industries 11% Residential
50% 11%
Direct GHG
contribu�on of our
surveyed sectors to Process
the world GHG
14%
industries
emissions 13% Agriculture related
2% emissions
2%
17% Forest land
10%
Energy and utilities Aviation
Others (Landfills, ships, croplands etc)

Source: Capgemini Research Institute analysis. Data from CAIT Climate Data Explorer via. Climate Watch/Our World in Data,
accessed October 2020.

5
“Our assessment shows that with digitalization, artificial
intelligence, big data analysis, deep learning, we can realize
a green economy much more easily with those technologies
than without them,”

Dirk Messner,
President, German Environment Agency
Board Member, Stockholm Energy Institute.

Not surprisingly, organizations are under pressure to These are challenging goals for the world economy and
act. The October 2018 special report from the IPCC society, but technology offers an exciting opportunity for
(Intergovernmental Panel on Climate Change) stressed the organizations to make a real difference. “Our assessment
need to avoid global warming reaching 1.5°C more than pre- shows that with digitalization, artificial intelligence, big data
industrial levels. Similarly, the Paris Agreement, which was analysis, deep learning, we can realize a green economy much
adopted by 195 nations in December 2015, set that target at more easily with those technologies than without them,” says
2°C more than pre-industrial levels.7 Dirk Messner, president, German Environment Agency;
Board Member, Stockholm Energy Institute. “However, despite
Pressure from climate groups, regulators and shareholders
the potential, during the last two decades, these technologies
is mounting too. Kelly Levin, senior associate at the World
have not been used enough to really solve the climate and
Resources Institute, says: “There is a huge range of actions
environmental and earth system challenges. We still have non-
that are building pressure. For example, the movements that are
sustainable growth patterns in most economies and organizations
taking to the streets and calling for different actors to increase
around the world. We need to make the link proactively between
their ambition towards reducing emissions, to shareholders and
climate politics, earth system stabilization strategies on the one
corporate boards. Some of the highest emitting companies don’t
hand side and these modern technologies on the other.”
have science-based targets; so while organizations are taking
steps, it’s certainly not at the pace that we need to be consistent If organizations and countries globally do not take decisive
with the science.” action, our planet is set to warm by almost 30C by the end
of the century – double the rate needed to constrain the
Investors are now actively shunning companies with low
worst impacts of climate change.9 Over the last few decades,
ratings or performance on ESG metrics. One of the largest
organizations have been taking significant measures
asset managers in the US, BlackRock, recently put 244
such as decommissioning coal plants and directing more
companies “on watch” for failing to take sufficient action on
funds to renewables and less to fossil fuels. However,
climate change. It also took voting action (holding directors
using technology levers to contain the effects of climate
accountable or supporting shareholder proposals) against 53
change is rarely seen as the biggest enabler. Among the
of these.8 As a result, organizations across industries need
technology levers, while the spotlight often falls on point
to adopt radical methods, make adequate investments, and
solutions that address a specific outcome – such as carbon
deploy appropriate tools, technologies, and strategies to
capture technologies to remove CO2 from the atmosphere,
achieve this goal.

6 AI in Climate Change
or renewable sources of energy – few technologies offer from such outcomes, Google declared that it would no longer
advantages that cut across sectors and value chains. Recent build AI tools for oil and gas extraction.13 While it is important
advances in Artificial intelligence (AI) for instance, in image to acknowledge that AI applications have a GHG footprint,
recognition powered by deep-learning neural networks is as some of the recent academic research has shown, it is
one such fast-emerging field, and points to AI’s significant also important to put it into the right perspective.14 For
potential that is yet untapped. We believe that AI solutions most large businesses and public sector organizations, the
when deployed sustainably, complementing technological footprint of individual AI applications is miniscule compared
and various other levers to reduce carbon footprint, offer a to their overall climate footprint. But, as the use of AI gains
sizeable and promising opportunity that is worth exploring. wider adoption, organizations do need to ask themselves
how AI can be used sustainably and responsibly. On this
AI offers unique opportunities to accelerate
front, we believe that optimizing AI training and execution
organizations’ climate action
can significantly limit its climate impact, when designed and
Our recent research shows that the adoption of AI across deployed responsibly. For example, as an early adopter of AI
organizations has been on the rise with more than one for tackling climate change, Google was able to reduce the
in two organizations (53%) moving beyond pilots or amount of energy it used in its data centers by 15% by using
proofs-of-concept in a few or more use cases.10 Given that inputs from machine learning algorithms.15 Using the wealth
organizations are already beginning to deploy AI at scale, it is of data from usage of servers in data centers, Google was
an opportunity for organizations to also explore AI-enabled able to more efficiently regulate their cooling. Ultimately, this
use cases from a climate action perspective and how existing led to significant power savings. We explore more dimensions
AI deployments or new innovations could also bring climate- of this aspect in the final section of the report.
positive outcomes (see section 1 on “AI offers many climate
In terms of sectors, we focused on asset-heavy and product-
action use cases”).
based sectors in this research which have a sizeable emissions
In recent years, several AI use cases have been identified footprint, such as automotive, manufacturing, consumer
that offer significant value in improving energy efficiency, products & retail, and energy & utilities sectors. Services
reducing dependence on fossil fuels, and optimizing sectors such as banking, insurance, and public services are
processes to aid productivity – all contributing significantly beyond the scope of this research. We do acknowledge that
to climate action.11 For instance, Air France uses Sky Breathe these services sectors cut across the industrial landscape and
– an ML and AI platform to provide a series of recommended do play a critical role from a financial, risk management, and
actions that can reduce the total fuel consumption by up to regulatory standpoint, among others. Achieving international
5%. Various airlines have used the solution to save more than climate goals will require political, social, and economic
590,000 tons of CO2 in 2019.12 Getting the focus right can willpower and action besides technological advances in many
have significant benefits in the long run. As we heard from areas including AI – the core focus of this report.
James McCall, Senior Director Global Climate and Supply
Based on our research and analysis, this report looks deeper
Chain Sustainability, Procter & Gamble “AI has the potential
into the following areas:
unlock a lot of data capabilities that we don’t have today, such
as predictive energy analytics at the production line level. Are 1. The most impactful AI use cases for climate action and
those going to change your global footprint overnight? No, they where organizations stand in adopting them
are going to typically be incremental, but changes to energy 2. The potential that AI offers for accelerating climate
efficiency – such as driving a 4% to 5% energy reduction across action and the benefits that have been achieved so far
our global manufacturing sites – is worth millions of dollars while 3. What a group of high-performing organizations can teach
helping significantly reduce our GHG emissions.” us about how to align and leverage AI capabilities to
execute on and achieve climate goals
While AI offers an exciting new front in the fight against
4. How organizations can leverage the full potential of AI
climate change, organizations have to be cognizant of AI’s
for climate action.
own climate footprint. Training AI models can contribute
to significant emissions and AI could be used to further
operations that accelerate climate change. In a move away

USD 2.97 trillion

Global economic losses due to climate
change from 2000-2019 as per the UN

7
 Artificial Intelligence

AI APPLICATIONS

Speech Image
Synthesis Processing
INTELLIGENCE
CAPABILITIES
e M
Conversational
ag Audio & Video
ces gu

LEARNING

ac Sens
Agents Processing
g
Pro Lan

hin ing
&
SYSTEMS
sin

eV
ral

ision
Natu

Natural Deep Reinforcement Analytics &

Language Learning Learning Predictions
Understanding

AI Acceleration Big Data

Intelligent Hardware Systems Data
d

Process Augmentation
an
Ac

ti AI ENABLERS
g

Automation
g
to n g a
in

tin
Au

ak

m nd d i c M
at
in g P re ion
i s
Autonomous D ec Advanced
Vehicles Simulations

Augmented Prescriptive
Creativity Modelling
AutoML

Source: Capgemini Technology, Innovation & Ventures.

Artificial Intelligence (AI) is a collective term for the capabilities shown by learning systems that are perceived by
humans as representing intelligence.
These intelligent capabilities typically can be categorized into machine vision and sensing, natural language
processing, predicting and decision-making, and acting and automating.
Various applications of AI include speech, image, audio and video processing, autonomous vehicles, natural
language understanding and generation, conversational agents, prescriptive modelling, augmented creativity,
intelligent process automation, advanced simulations, as well as complex analytics and predictions.
Technologies that enable these applications include big data systems, deep learning, reinforcement learning, and
AI acceleration hardware.

8 AI in Climate Change
1. AI offers many climate action
use cases
AI offers a number of use cases that positively affect the the fight against methane leakages) as well as those that
climate, both in preventive measures such as reducing GHG can contribute to positive climate outcomes (such as AI for
emissions and remedial measures such as dealing with the general energy efficiency). Figure 2 summarizes them.
effects of climate change more effectively. As we heard
In addition, large insurers and reinsurers are using AI for
from David Rolnick, chair, Climate Change AI and assistant
climate and environmental risk modeling and helping
professor of Computer Science at McGill University, says: “In
their clients with it. Munich RE, a provider of reinsurance,
both power generation and distribution, there are many ways
primary insurance and insurance-related risk solutions, is
in which machine learning can be used to increase efficiency,
experimenting with AI in its risk solutions. As we heard from
support low-carbon electricity generation and, in general, help
Christof Reinert, head of Risk Management Partners at
reduce greenhouse gas emissions. Likewise, in traditional industry
Munich RE, “We have a Risk Suite which is the umbrella for all
and manufacturing – both in the manufacturing component
our risk solutions we offer as a cloud-service. The comprehensive
and in the distribution of goods – there is the capacity for
set of solutions cover location risk and data risk intelligence,
vastly improved efficiency and lower energy use using machine
in both of which we have the ability to use Munich Re’s smart
learning.” The use cases include both AI systems that are
company matching algorithms which are trained on AI models.”
designed to specifically address climate challenges (such as

“In both power generation and distribution, there are many ways in which machine learning can
be used to increase efficiency, support low-carbon electricity generation and, in general, help
reduce greenhouse gas emissions. Likewise, in traditional industry and manufacturing – both in
the manufacturing component and in the distribution of goods – there is the capacity for vastly
improved efficiency and lower energy use using machine learning.”

David Rolnick,
Chair, Climate Change AI and assistant professor of Computer Science at McGill University

9
Figure 2. Using AI for climate action across sectors and value chain
Energy production and distribution

Tracking emissions at industrial sites Energy efficiency

Tracing leaks using smart leakage AI can help improve energy efficiency and
management to avoid GHG emissions use cleaner electricity sources when available
GHGSat uses machine learning to look for potential
areas of methane leakages and autonomously Google’s Deepmind uses machine learning to
monitors such areas from the satellite based predict wind patterns at their 700 MW wind
methane detectors.(i) farm upto 36 hours in advance leading to
closer alignment with the grid.(iii)
ExxonMobil has teamed up with MIT to design and
develop deep-sea AI robots to boost their oil seep
detection capabilities. Oil seep could account for nearly
half of the oil released into the ocean every year.(ii)
Manufacturing operations

Route optimization and fleet management Maintenance

AI for route and fleet management can help reduce AI can help predict and avoid manufacturing
fuel consumption and carbon emissions disruptions due to mechanical failures

Rio Tinto uses machine learning and sensors to track General Motors used image classification
vital vehicle stats and asset efficiency, leading to tool on nearly 7,000 robots to detect
increase in serviceable life.(vi) component failure leading to more efficient
operations and lower energy costs.(viii)
Locus.sh uses AI for route optimization for last mile
deliveries to achieve 25% efficiency gain and GHG
reductions.(vii)

Inventory management
Product usage, reuse, and recycling

AI for inventory management to improve store stock and

to closely align demand and supply
PepsiCo is using big data and advanced analytics in Pep Worx to help identify potential households and
avenues of sale, leading to more accurate demand forecasting and lower waste.(xi)

Morrisons uses AI to optimize demand and replenishment at a store and product level using internal datasets
(e.g., sales data) and external data (e.g., weather forecast) leading to reduced wastage.(xii)

Wasteless, an Israeli startup, uses AI to dynamically price perishable products on retail shelves leading to
wastage reduction of 39% in a 12 week pilot.(xiii)

Energy Utilities Automotive Industrial manufacturing Consumer products Retail

10 AI in Climate Change
 Energy production and distribution
Dynamic peak load adaptation
AI technologies can help manage peak demand planning and thus avoid energy wastage

EnPowered uses AI to adapt to demand response systems, thereby reducing peak load,
leading to closer alignment of demand and supply and reduction in energy wastage.(iv)

Robert Bosch GmbH uses AI to predict future energy consumption, avoiding high peaking
loads, and manage deviations in patterns of consumption resulting in emissions reduction
by 10% in two years.(v)

Manufacturing operations
Operational effectiveness New product design
AI can help increase operational effectiveness AI for designing new products to build more
to increase efficiency and better utilization environment-friendly products and reduce
of resources wastage in design

Baker Hughes uses AI to provide operators and General Motors used machine learning to
engineers insights into machine health and real build a new seatbelt bracket design that is
time control of production processes leading to 40% lighter and 20% stronger with reduced
12% better machine utilization.(ix) resource wastage during design.(x)

Consumer behaviour analytics Waste management & segregation

Product usage, reuse, and recycling

AI for consumer behaviour analytics AI for effective waste management,

to reduce wastage segregation and recycling to reduce need of
resources & avoid wastage
Bidgely uses machine learning algorithms and
database of more than 50 billion meter Stuffstr uses AI algorithms for pricing and
readings to identify down to the appliance level discovering the value of reused items. They also
the most energy consuming devices and usage use machine learning to ensure proper
patterns in households giving deeper visibility classification of all re-sale products.(xv)
of decarbonization opportunities.(xiv)
ZenRobotics uses AI to assess waste stream in
real time and separate waste quickly and with
high precision resulting in effective recycling.(xvi)

11
 Sources for Figure 2:

i. PRNewswire, “Meet Iris: GHGSat to Launch a Second Emissions-Monitoring Satellite This Summer,” April 2019.
ii. SparkCognition, “Top 4 Real-World Ai Applications In The Oil And Gas Industry”, March 2020
iii. Deepmind, “Machine learning can boost the value of wind energy,” February 2019.
iv. Startup Toronto, “EnPowered Secures Stake in Ontario’s Demand Response Electricity Program, Aiming to Improve Customer
Revenue,” December 2019
v. Bosch press release, “CO₂ advisory service: Bosch assists manufacturing companies as they work toward climate neutrality,”
July 2019.
vi. IoTforall, “Go Big on IoT Value with Preventive Maintenance,” July 2017
vii. Locus.sh, “How the World’s Best Route Optimization Engine Works,” January 2018.
viii. iFlexion, “Industries to Be Transformed by Machine Learning for Image Classification”, October 2018
ix. Capgemini, Artificial Intelligence (AI)-Driven Smart Factory solution provides operators and engineers with a new level of
insight and the ability to adjust production at a moment’s notice, June 2019
x. Autodesk, “Automotive Lightweighting with Generative Design.”
xi. Forbes, “The Fascinating Ways PepsiCo Uses Artificial Intelligence and Machine Learning to Deliver Success,” April 2019
xii. Retail week, “How machine learning is enabling Morrisons to innovate around its customers,” March 2017
xiii. Singularity hub, “Food Waste Is a Serious Problem. AI Is Trying to Solve It,” November 2020
xiv. Businesswire, “Bidgely Surpasses 10 Million Homes Under Contract With Utilities for Energy Disaggregation”, December 2017.
xv. StuffStr website, accessed September 2020
xvi. Recycling Product News, “ZenRobotics’ AI-based robotic waste sorting technologies help Remeo to build next-generation
MRF,” September 2020.

A range of organizations have already made innovative use of emissions. Using AI-powered augmented reality, field
AI to address the adverse impact of GHG emissions, drawing technicians are quickly able to link with remote technical
on different data, such as satellite images, drone footages, support and narrow down the emission sites.19 As we heard
and weather information: from Patrizia Radice, CHRO, SARAS Group, a large player
in the European refining industry, “We started digitalization
• Tracing methane leaks and emissions. Methane is one
of the refinery three years back. It was a large program with
of the most potent greenhouse gases – 84 times more
a significant investment. As part of this, AI was one of the
potent than carbon dioxide, it constitutes around a quarter
drivers to make better use of historical data to understand
of manmade global warming.16 Tracking and plugging
what is going on and support predictive maintenance of
methane leaks from oil & gas facilities, landfill sites, mines,
machines so that they offer greater energy efficiency and
and agriculture has been a daunting task for public and
reduce GHG emissions.” AI-powered data analytics platforms
private enterprises. GHGSat – a startup focused on using
are also helping monitor vital climate parameters such as
satellites to detect greenhouse gas leaks that would be
atmospheric concentration of CO2 to better understand its
highly difficult by other means on land – uses AI to offer
impact on climate.20
insights into emissions, potential areas of leakage, and
then autonomously monitors and alerts on any irregular • Optimizing renewable energy generation. Greensmith
activity.17 In January 2019, GHGSat helped locate a leak that Energy, a global energy storage company, uses machine
is estimated to have released 142,000 tons of methane learning to manage energy storage systems and balance
in 12 months prior to discovery.18 In the US, the British multiple energy assets such as wind, solar, and diesel
multinational oil and gas company, BP, has used drones generation. The new, AI-powered “Graciosa Hybrid
to track 1,500 well sites to monitor and trace methane

12 AI in Climate Change
 Renewable Power Plant,” automatically optimizes energy – Air France uses Sky Breathe – an ML and AI platform
generation based on energy demand and weather patterns to suggest a series of recommended actions that can
(that affect wind and solar power generation). It improves reduce the total fuel consumption by up to 5%. Various
reliability of the energy grid while aiding larger adoption of airlines have used the solution to save more than
renewable power from 15% to 65% and reducing the need 590,000 tons of CO2 in 2019.24
for 17,000 liters of diesel per month.21
– Alibaba uses real time data such as GPS, traffic
• Reducing wildfire risk and adapting operations to conditions, and AI to make routing decisions for its
changing weather. Utilities in California are exploring the last-mile delivery fleet. These systems were able to
use of drones, along with computer vision, to regularly reduce distance travelled for the fleet by 30% and
monitor transmission and distribution equipment for reduce vehicles needed by 10% in a pilot run.25
faults. This is part of efforts to reduce their risk of starting
– Ocado – the British online grocery retailer – has
fires. Apart from being a threat to lives and livelihood,
deployed machine learning-powered forecasting and
wildfires also contribute heavily to carbon emissions. The
optimization to reduce food waste. Ocado has been
California wildfires this year have already generated 91
able to slash food wastage rates to just 1 in 6,000 items
million metric tons of carbon dioxide as of September,
by using data analytics, machine learning and artificial
25% more than the state’s annual fossil fuel emissions.22
intelligence to manage its produce.26 Reducing food
PG&E – a utility in California – has been using drones with
waste helps avoid methane build-up – a greenhouse gas
computer vision for remote aerial inspection, and analytics
much more potent than CO2.
and machine learning to predict how its transmission
equipment will handle high-wind events. This allows the
organization to prioritize maintenance work on vulnerable
parts of the transmission chain.23

“We started digitalization of the refinery three years back. It was a large program with a
significant investment. As part of this, AI was one of the drivers to make better use of historical
data to understand what is going on and support predictive maintenance of machines so that
they offer greater energy efficiency and reduce GHG emissions.”

Patrizia Radice,
CHRO, SARAS Group

13
2. AI-enabled use cases are
already reducing GHG emissions
and have the potential to
accelerate climate action
AI-enabled use cases have helped reduce GHG emissions by 13% since 2017

AI has been helping organizations around the world as This is a significant boost to organizations’ climate actions and
they tackle the impact of climate change on their business goals to achieve carbon neutrality and net-zero emissions.
operations and achieve their sustainability goals:
Our survey found that, in the last two years, organizations
• Nearly half of the organizations (48%) we surveyed have have been able to reduce their GHG footprint by 13% using
used AI for climate action. AI-enabled use cases.
• As a result of AI, organizations have also reduced GHG We analyzed the average benefits yielded by AI use cases for
emissions by 12.9%, improved power efficiency by 10.9%, climate action when deployed at partial scale or full scale (see
and reduced waste by 11.7% since 2017 (see Figure 3). Figure 3). Overall, these projects were able to deliver 11–13%
gain across various KPIs in the last two years.

Figure 3. AI-enabled use cases have helped organizations achieve several sustainability benefits, including reducing
GHG emissions by 13% in two years

Average benefits realized from using AI for climate action

12.9% 11.7% 11.5%

10.9%

Reduced Improved Reduction of waste and Cost savings from

greenhouse gas power/industrial dead weight assets climate mitigation
emissions efficiency

Achieved in the last two years (with base year 2017)

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=190 organizations that have been able to fully scale
or partially scale AI projects for climate action. Reduction of waste includes waste in broader terms of utilization deadweight – empty trucks/
facilities and in disuse/disposal of products before completion of useful life (e.g., produce/vehicles).

14 AI in Climate Change
Utilities and consumer products organizations have experienced the biggest gains in reducing GHG emissions using AI
(see Figure 4).

Figure 4. Utilities and consumer products have realized the biggest GHG reduction using AI-enabled use cases

Average GHG emission reduction though AI-enabled use cases in the last two years-by sector
(base year 2017)

13.9% 14.3%
13.5% 13.3%
12.6% 12.9%
12.3%
11.3%

Utilities Consumer Medical Retail Automotive Fossil Fuel Others Global

(Electricity, Products Devices (Oil,Coal)
Gas,Water)

Source: Capgemini Research Institute, AI in climate action survey, July-August 2020, N=190 organizations that have been able to fully or partially
scale AI projects for climate action. Others include process industry (cement, paper, petro-chemical, paper) and discrete industries (electrical
and electronics, air and railway equipment, etc.)

14%
average GHG emissions reduction attained by
the utilities sector from AI-enable use cases.

15
AI-enabled use cases have the potential to significantly limit GHG emissions

"We tend to overestimate the effect of a technology in the short the Copenhagen Centre on Energy Efficiency, said, “There
run and underestimate the effect in the long run” - Roy Amara, are hundreds of very interesting opportunities that are coming
former president of The Institute for the Future.27 up based on the use of AI. I truly think that there are huge
opportunities for energy systems at a household, municipal, or
Our survey shows that organizations expect to cut GHG
even country level that could improve their performance with the
emissions by 16% in the next 3–5 years through AI-driven
aim of reducing the energy demands associated with providing
climate action projects (see Figure 5). Speaking about the
the same level of services.”
future of AI for climate change, Gabriela Prata Dias, head of

Figure 5.1. AI has the potential to deliver considerable climate gains across sectors

Average benefits expected from use of AI-enabled use cases climate action in the next three to five years

15.9% 14.7% 15.7% 14.2%

Reduced green house Improved power/industrial Reduction of waste and Cost savings from
gas emissions efficiency dead weight assets climate mitigation

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=190 organizations that have been able to fully scale or
partially scale AI projects for climate action. Reduction of waste includes wastage in broader terms of utilization deadweight – empty trucks/
facilities and in disuse/disposal of products before completion of useful life (e.g., produce/vehicles).

Figure 5.2. AI has the potential to deliver considerable climate gains across sectors

Average future emission reduction using AI-enabled use cases for the next three to five years
(base year 2019)
18.3% 17.6%
16.5% 15.9% 15.5% 15.2% 15.0% 15.9%

Medical Energy Utilities Consumer Automotive Retail Others Global

Devices (Oil,Coal) (Electricity, Products
Gas,Water)

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=190 organizations that have been able to scale AI
projects fully or partially for climate. Others include process industry (cement, paper, petro-chemical, paper) and discrete industries (electrical
and electronics, air and railway equipment, etc.).

16 AI in Climate Change
While this is a significant reduction in GHG emissions, would it To explore whether GHG emission reduction through AI
be sufficient to get anywhere close to international emission may help companies to align their climate impact to a global
reduction agreements such as the Paris Agreement? For warming well below 2°C, we partnered with a climate change
instance, in line with the Paris Agreement, the EU has set a start-up – ‘right. based on science’ – and used their X-Degree
target of reducing its emissions by at least 40% by 2030 and is Compatibility (XDC) Model (see insert on “What is the XDC
considering increasing it to 60% reduction against Model and how to interpret it?" and Appendix “The XDC
1990 levels.28 Model” for further information).

What is the XDC Model and how to interpret it?

The XDC Model calculates the contributions of a – a relationship between the emissions produced
company, portfolio or any other economic entity per generation of one million euro Gross Value
to climate change, answering the question: How Added (GVA).29 The EEI shows the ability of an
much global warming could we expect, if the entire organization to decouple their economic growth
world operated at the same economic emission from their emissions. A growing EEI indicates that
intensity as the entity in question? Results are the organization’s emissions outpace its economic
expressed in a tangible degree Celsius (°C) number: growth which is unsustainable and harmful for the
the XDC. The XDC Model, a science-based and peer- climate. Ideally, organizations’ EEI must reduce with
reviewed methodology, is the only one of its kind time as they reduce their GHG emissions or grow
to integrate a full climate model (also used by the them slowly while growing their business. This
UN Intergovernmental Panel on Climate Change, or reduction in EEI must be more than that intended by
IPCC). international targets (for instance, by 2030, 45% of
2018 levels for most sectors).
The main input parameter for the XDC Model is
a metric called Economic Emission Intensity (EEI)

“There are hundreds of very interesting opportunities that are coming up based on the use of
AI. I truly think that there are huge opportunities for energy systems at a household, municipal,
or even country level that could improve their performance with the aim of reducing the energy
demands associated with providing the same level of services.”

Gabriela Prata Dias,

Head of the Copenhagen Centre on Energy Efficiency

17
Using the XDC model and its input parameter – the Economic sector-specific estimates from our survey are used (the
Emission Intensity (EEI) – it is possible to assess at what detailed methodology is described in the Appendix). For
rate the GHG emissions of organizations and sectors will the automotive sector example, as shown in Figure 6,
evolve over time in various scenarios. To model the effect this scenario represents a total EEI improvement of 8% in
of AI-enabled climate use cases on organizations’ EEI – i.e. 2030 over the Baseline Scenario. This means that by using
quantifying the extent to which AI help reduce EEI – we use AI-enabled use cases for emission reduction, automotive
two scenarios (see Appendix B for a more detailed view of organizations can meet more than a fifth of their EEI
the scenarios): reduction requirement by 2030.
1. Baseline Scenario: here, the GVA and the sector’s
We then compare these two scenarios with the “Beyond 2
emissions continue to grow in line with historic trends
Degree Scenario” (B2DS) of the International Energy Agency
with no planned emission reduction initiatives, such as
– a scenario that aims to limit global warming to 1.75°C,30 in
using AI to curtail emissions. As the net effect of this, the
line with the Paris Agreement. Using the XDC Model and the
EEI of organizations reduces gradually. This is in line with
emissions budgets from the B2DS, we compute an EEI target
the “Shared Socioeconomic Pathway 2” (SSP2) scenario.
pathway for each sector. This EEI target pathway represents
how fast organizations’ EEI must reduce in this sector to be
2. “AI-enabled” Scenario: here, the sector implements
compliant with the Paris Agreement. For instance, the target
AI technologies and delivers the resulting GHG emission
pathway for the automotive sector requires an EEI reduction
reduction benefits. GVA growth is assumed to follow
of 37% by 2030 over Baseline levels (see Figure 6).
historic trends; for the GHG emissions reduction rates,

Figure 6. AI-enabled use cases have the potential to assist automotive organizations to meet more than one-fifth of
their Paris Agreement emission goals by 2030

Evolution of Economic Emissions Intensity Pathways for the automotive sector

2800
(tCO2e/Mio EUR Gross Value Added)

2600
Economic Emissions Intensity

2400
2200
2000
8%
1800
1600
1400
1200 37% 45%
1000
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

B2DS pathway (for limiting global warming to 1.750C per IEA and Paris Agreement)
“AI-enabled” pathway
Baseline (business-as-usual) pathway

Source: Capgemini Research Institute and right. based on science analysis.

The Economic Emissions Intensity of an organization or a sector is the amount of emissions produced by it per generation of one million euro
Gross Value Added (GVA). The EEI must ideally reduce with time as organization/sector grows while reducing its emissions. This reduction
compared to 2018 levels is represented by EEI pathways. The B2DS Pathway represents EEI reduction aligned to the Paris Agreement goal
of containing global temperature increase to under 20C (1.750C for B2DS). The Baseline Pathway assumes that GVA and emissions continue
to grow as per historical trend. As the net effect of this, EEI reduces gradually. The AI-enabled Pathway (optimistic adoption) represents the
reduction in EEI when organizations adopt AI-enabled use cases and derive their benefits on GHG emission reduction as observed
in our research.

18 AI in Climate Change
We conducted the above analysis for five sectors in our can deliver 11–45% of the requirement between 2018–30.
survey and for each sector, and we found that the use of Consumer retail has the greatest potential for AI-driven
AI-enabled use cases yields a 4–17% of EEI reduction by 2030 improvements at 45%, while wholesale retail offers the least
compared to the Baseline Scenario (see Figure 7). In terms at 11%.
of the share of target EEI reduction, AI-enabled use cases

Figure 7. AI-enabled use cases have the potential to aid organizations to achieve 11-45% of their EEI reduction targets
by 2030

AI’s potential AI's potential in Economic Emissions Intensity reduction

contribution as share by 2030 compared to baseline reduction
of EEI reduction

36%
18% Oil and Gas
7%

37%
22% Automotive
8%

38%
11% Wholesale retail
4%

38%
45% Consumer retail
17%

42%
32% Utilities
14%

B2DS pathway reduction (for global warming to 1.75⁰C per IEA and Paris Agreement)
AI-enabled pathway reduction

Source: Capgemini Research Institute and right. based on science analysis.

19
45%
The potential of AI's contribution to reach
the required goal of IEA's Beyond 20C
Scenario for the consumer retail sector.

The COVID-19 crisis has slowed organizations’ progress on climate goals

The COVID-19 crisis has made people more aware of However, in the immediate term, we found that over
sustainability issues. It has also led to some positive a third of sustainability executives (37%) said that the
policy initiatives, especially in Europe. The European COVID-19 crisis has decelerated their climate goals
Parliament has put a “Green Deal” at the core of its (see Figure 8). The deceleration is at the highest in
economic recovery from COVID and it aims for the EU the energy and utilities industry. As Dr. Faith Barol,
to be climate neutral by 2050. 31 In October 2020, the executive director of the International Environmental
European Union parliament voted to cut greenhouse Agency noted, “Despite a record drop in global
gas emissions by 60% by 2030 compared to emissions this year (2020), the world is far from doing
1990 levels.32 enough to put them into decisive decline.” 33

Figure 8. Impact of COVID-19 pandemic on progress against climate goals: accelerated vs. decelerated

How has the COVID-19 crisis affected your long-term climate goals?

37% 39% 41%

36%
33%

20%
14% 14% 11% 11%

Overall Manufacturing/ Energy and Utilities Automotive Consumer Products

Process industry and Retail

It has accelerated our actions towards our climate goals

It has decelerated progress on our climate goals and pushed them further away

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=400 sustainability executives.

20 AI in Climate Change
Looking deeper into potential reasons behind this • 38% have put a hold on capital expenditure
deceleration, and as Figure 9 shows, we found that: allocated for climate initiatives (increases to 47%
in process manufacturing).
• 46% said they had put one or more climate
initiatives on hold (this increases to 53% in
automotive).

Figure 9. One in two organizations have put climate initiatives on hold owing to the COVID-19 crisis

Which of the following is true for your organization owing to the COVID-19 crisis?

53%
46% 49% 50% 46%
42%
38% 41% 36% 32%
40% 38%
37%
41%
38%

Overall Automotive Consumer Products Energy and Utilities Manufacturing/

and Retail Process industry

We have had to put one or more climate change initiatives on hold

We have had to put capital expenditure for some or most climate initiatives on hold
We expect a loss of capabilities and talent critical for our climate change initiatives

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=400 sustainability executives.

41%
Share of organizations that expect a
loss of capabilities and talent critical for
climate change initiatives due to Covid-19.

21
3. Even though climate action is a
strategic priority, most
organizations are struggling to
support climate action with
AI capabilities
Climate action is a top priority for organizations

Climate action and controlling emissions are major strategic • 70% of sustainability executives said they have a
priorities for organizations and their leaders: governance body to oversee their organization’s climate
objectives and/or review the progress achieved by the
• 67% have made long-term business decisions to tackle the
sustainability/climate change team.
consequences of climate change.
As Figure 10 shows, the most popular areas of focus are
• 60% of sustainability executives said that climate change
driving carbon-neutral operations, focusing on achieving net-
strategy is embedded in their business strategy.
zero greenhouse gas emissions.

Figure 10. Top five actions that form a part of organizations’ climate action

1 Carbon-neutral operations: Carbon emissions balanced by carbon savings

2 Paris Agreement temperature alignment: Contributing to limiting temperature increase

3 RE-100: Shifting to 100% renewable energy for all operations and energy use

4 Emission reduction: Setting goals for science-based overall reduction of GHG emissions from operations

5 Industry benchmark: Benchmarking organization’s climate progress vis-à-vis its industry

Source: Capgemini Research Institute, AI in climate action survey, July–August 2020, N=400 sustainability executives.

22 AI in Climate Change
As they set their ambition and goals, half (52%) have aligned we will continue to see technology becoming cheaper but
their chosen approach and actions to the wider United carbon credits are likely to get costlier (as current price
Nations’ Sustainable Development Goals (SDG) agenda, and levels are substantially lower than the price levels deemed
37% plan to do so the future. However, in terms of their consistent with achieving the temperature goal of the
long-term strategy, 84% executives said that they would Paris Agreement).34 This approach points to the fact that
rather compensate for (or offset) their carbon footprint than organizations may not be looking at the issue holistically and
deploy technology solutions to reduce their footprint (16%). with a long-term view.
This points to a less than optimal approach. This is because

Few organizations are successfully combining their climate vision with AI capabilities

As we have shown above, AI can play a substantial role release are affecting the climate. There is also the concern about
in ensuring organizations meet Paris Agreement goals. how companies might be affected by climate change and how
However, this would require having a cohesive climate their facilities might be/are being disrupted from major storms
strategy backed by AI capabilities. Our analysis shows that or how the supply chains are being disrupted. They are taking
only a small minority of organizations has what it requires various measures to reduce their emissions, although they are
to successfully leverage AI solutions for climate action. As obviously concerned about underlying profitability. It will be
we heard from John Reilly, co-director, MIT Joint Program critical for organizations to think innovatively including with
on the Science and Policy of Global Change, “By this time the use of emerging technologies such as AI, to harness their
industries are aware that greenhouse gas emissions that they climate actions.” We call this leading group the “Climate AI
Champions” (see “How do we classify Climate AI Champions?”
for a more detailed explanation of how we define
this cohort).

“By this time industries are aware that greenhouse gas emissions that they release are affecting
the climate. There is also the concern about how companies might be affected by climate change
and how their facilities might be/are being disrupted from major storms or how the supply
chains are being disrupted. They are taking various measures to reduce their emissions, although
they are obviously concerned about underlying profitability. It will be critical for organizations
to think innovatively including with the use of emerging technologies such as AI, to harness their
climate actions.”

John Reilly
Co-director, MIT Joint Program on the Science and Policy of Global Change

13%
Share of organizations that have been
able to couple strong AI applicabilities
with concrete climate action and strategy.

23
 How do we classify Climate AI Champions?

In our analysis, we found very few organizations are therefore in pole position to turn AI’s climate
that have aligned their climate vision and strategy potential into action and value, we analyzed
with their AI capabilities. To understand which all surveyed organizations based on these two
organizations have achieved alignment, and which dimensions:

1. Climate action vision and governance 2. AI capabilities and execution

Climate vision AI capabilities/talent

This covers whether an organization has an We assess organizations on whether leadership and
overarching climate strategy along with the employees are willing and able to use data for
relevant goals they are pursing climate action

Execution and governance Ability to scale AI use cases

We evaluate whether an organization has We evaluate organizations based on their success to

embedded their vision with their business strategy date in scaling AI use cases for climate action
and look at their sustainability/climate change
governance structure

Climate action alignment AI investment and propensity

We look at their priorities and whether execution We assess their appetite for using AI along with
effort is focused on the areas of their value chain their committed budgets/investment
responsible for the greatest amount of
carbon emissions.

We found four broad groups of organizations within our 400 surveyed organizations (see Figure 11):

AI Advocates Climate AI Champions

They have strong AI skills but lack climate They have a mature climate change vision,
change governance, vision and execution strategy, and strong record of accomplishment of
(26% of organizations). AI implementation for climate action. They
constitute 13% of all surveyed organizations.

Climate AI Laggards Climate Advocates

They lack both climate change vision and While they have a strong climate change vision
execution capability and do not possess the and execution capability, they lack the AI
required AI capabilities (44% of organizations). capabilities to deliver (18% of organizations).

24 AI in Climate Change
Figure 11. Only 13% of organizations combine climate vision and execution with AI capabilities

AI/Climate change maturity matrix

Climate advocates—18% Climate AI Champions —13%

Climate action visions and governance

Climate AI Laggards —44% AI advocates —26%

AI capabilities and execution

Source: Capgemini Research Institute, AI in climate change survey, July–August 2020, N=400 organizations.

Four in ten of our climate AI champions come from the EU, sectoral analysis shows an equal distribution of Climate AI
followed by the Americas and APAC (see Figure 12). Our Champions among all our four sectors.

Figure 12: Climate AI Champions mostly hail from Europe

Geographical distribution of Climate AI Champions

Europe 42%

30% Americas

Asia-Pacific 28%

Source: Capgemini Research Institute, AI in climate change survey, July–August 2020, N=51 Climate AI Champions.

25
 Lack of investments and skills is holding back both experimentation and
the drive to scale in climate AI
Despite the considerable potential for AI for climate • While organizations have evaluated 88% of
action, our research shows that this potential is them for implementation, only 42% have been
largely untapped in terms of either experimentation prototyped/piloted.
or scaled deployment. As Figure 13 shows, while • 11% of all use case have been partially scaled and
awareness is strong, far fewer have experimented only 3% of all AI use cases have been fully scaled.
with – or deployed – the 70 use cases we evaluated:

Figure 13. 88% of AI use cases for climate action are evaluated but less than half of them are
experimented with

Percentage of AI use cases for climate change that...

Apply to their respective industry 100%

Evaluated for implementation 88%

Prototyped/piloted stage 42%

Partially-scaled
11%
(at one or multiple sites)

Fully-scaled (deployed in
3%
production and continued scaling more
throughout multiple business teams)

Source: Capgemini Research Institute, AI in climate change survey, July–August 2020, N=400 organizations.

Our survey also reveals why there is a lack of • Half of all organization (54%) have fewer than 5%
progress from evaluation of use cases (88%) to the of employees with the skills to take up data and
pilot/prototyped stage (42%): AI-driven roles.
• More than eight out of 10 organizations spend less
These challenges closely echo the findings from our
than 5% of climate change investment towards AI
other research on scaling enterprise AI.35
and data tracking.

26 AI in Climate Change
Climate AI Champions can take more cohesive climate action with the help of AI

To ascertain how well placed Climate AI Champions Agreement. We have further broken down the resulting
are to meet their climate goals vis-à-vis the rest of the Portfolio XDC Gaps based on Scope 1 and 2 emissions (see
organizations, we analyzed the temperature alignment of Appendix D on “Emissions Accounting” for more details on
the Climate AI Champions with Paris Agreement goals and emission scopes).
compared it with the other companies in our research. To do
As Figure 14 shows, Climate AI Champions have significantly
so, we have analyzed the individual companies in our survey
lower Portfolio XDC Gaps than their peers in both emission
and consolidated them on a group – or “portfolio” – level
1 and 2 scopes (see appendix “Emissions Accounting” for
according to the classification above. This analysis results in a
detailed scope definitions and “The XDC Model” for a
temperature alignment. The Portfolio XDC Gap signifies the
description of XDC metrics). This indicates that the ability to
gap that remains to be closed to meet the Paris Agreement
leverage AI in tandem with a concrete climate strategy brings
target of <20C warming (See Appendix AII on the “Portfolio
Climate AI Champions closer to the Paris Agreement goals.
XDC Metrics” for more details on this methodology.) A
gap of 00C represents complete alignment with the Paris

Figure 14. Climate AI Champions are closer to Paris-Alignment compared to their peers in both scope 1 and 2 emissions.

Portfolio XDC Gap for Climate AI Champions vs the rest of the organizations - the level of warming that
must be reduced by these group of companies to be aligned with the Paris Agreement

2.5°C

1.2°C

0.7°C 0.7°C

Scope 1 XDC Gap Scope 2 XDC Gap

Climate AI Champions The rest of the organizations

Source: Capgemini Research Institute, AI in climate change survey, July–August 2020, N=400 organizations; right. based on science analysis.
Data indicates the °C gap that organizations must close, in order to meet their sector-specific requirements in line with the Paris Agreement.
Scope 1 accounts for emissions that are from internal emissions and Scope 2 accounts from electricity usage from local grids and utilities. Check
appendix for emission scopes and XDC Gap definition.

0.70C
The gap for Climate AI Champions to achieve their Paris
Agreement goal for Scope 1 emissions, compared to
2.50C for the remaining organizations.

27
However, while Climate AI Champions are performing better Organizations will also need to combine the increased use
than their peers, both groups still have positive (>0) XDC of AI with further measures, e.g. stronger climate strategies,
Gaps, signifying that they are not yet aligned with the Paris financial incentives and investments, and risk management,
Agreement. The XDC Gap can only be closed by decoupling to close the gap further. For Climate AI Champions, the
emissions from economic growth, resulting in a reduction of next frontier is ensuring that gains from AI implementation
Economic Emission Intensity (EEI), for example through the make themselves felt across their supply chain and indirect
implementation of technology. This analysis demonstrates emissions. For the rest of the cohorts, AI can play a critical
the potential of strong AI capabilities in alignment with role in building carbon tracking, reporting and estimation.
a mature climate strategy to achieve a reduction in EEI,
indicated by a closing XDC Gap.

What makes Climate AI Champions stand out?

As discussed, and shown above, Climate AI Champions have Climate AI Champions are also able to drive their climate
made considerable gains in applying climate AI to their direct action and AI action by building dedicated leadership and
emissions. This is a result of them having their core operations technical teams to drive implementation:
as a focus of their climate strategy (see Figure 15):
• 47% have a dedicated leader for each climate goal and
• 60% of Climate AI Champions have internal core operations action compare to 39% of others.
in the scope of their climate strategy compared to 27% of
• 53% have a dedicated team for implementing tech
the rest.
solutions compared to 42% of others.
• 60% of Climate AI Champions have intra- and-inter logistics
Climate AI Champions also take into account the indirect
in the scope of their climate strategy compared to 49% of
emission contributions from their extended supply chain
the rest.
of partners and vendors: 58% set extensive disclosures and
goals for their supplier compared to 39% of the rest.

Figure 15. Climate AI Champions have taken more progressive and concrete steps towards climate action.

Percentage of organizations taking the following actions

60% 60% 58%

53%
49% 47%
39% 42% 39%

27%

Internal core Inter and intra Dedicated leader Dedicated team Set disclosures and
operations as a part logistics as a part for each climate for implementing goals for suppliers
of their climate oftheir climate goal/action technology
strategy strategy solution

Climate AI Champions The rest of the organizations

Source: Capgemini Research Institute, AI in climate change survey, July–August 2020, N=400 organizations.

28 AI in Climate Change
 Interview with Matthias Berninger, SVP Public
Affairs, Science & Sustainability, Bayer

How are you using AI for climate action at Bayer? look at AI is for outcome-based models. So, for
example, how can you transform a business that is
We acquired Climate Corp. roughly two years ago,
built on selling seeds and pesticides to a business
which is helping Bayer collect more and more data.
that sells a solution combined with a guaranteed
One of the ways is that we use sensor technology
outcome. Outcome-based models like these are
to better understand what happens in the soil. Soil
hugely important for sustainability and require deep
is a huge carbon sink if it’s leveraged in the right
integration of IT and artificial intelligence with what
way and it can also be a huge problem if it’s going
you’re trying to achieve.
the other way. So that’s a business where we have
a lot of data points that we are leveraging with The same is true in the biological space and if you
computing technologies and artificial intelligence look at the Bayer seed business. It’s basically a
applications. business that uses everything from GMO to gene
editing to genome analysis. The whole process
Do your AI and sustainability teams collaborate
of developing new seeds is a fully integrated
for climate outcomes?
business that uses all kinds of artificial intelligence.
Yes, we do. For example, the implementation of Computing power meeting biology is one of those
quantitative sustainability targets is only possible big trends that I believe will replace the classic
through a conversation and through a strategic discussion about artificial intelligence fairly quickly.
collaboration with the IT teams. An area where we

60%
Share of Climate AI Champions that focus on internal
core operations as a part of their climate strategy vs.
just 27% of the rest of the organizations.

29
4. Recommendations: How
organizations can leverage AI’s
full climate action potential
Sustainability and the determination to address climate sustainability into the business strategy has become a prerogative
change is an increasingly important element of performance for businesses in every sector to continue to attract investors,
and critical for an organization to meet its stakeholders’ customers as well as talented employees.”
expectations, including investors, consumers, employees and
It is critical, therefore, that organizations leverage AI’s
communities. To meet increasing societal and stakeholder
potential in the sustainability space. As Meghna Tare, chief
expectations, organizations are looking beyond just ‘profit’
sustainability officer, University of Texas at Arlington, said,
to a wider business ‘purpose’. “Companies with strong
“Tackling climate change has to be part of a strategic plan for
sustainability strategies significantly outperform the market
every organizations. Artificial intelligence (AI) appears naturally
– that is the conclusion of recent papers from some of the
poised to address transformational challenges of sustainability
world’s leading business schools.” says Tom Schalenbourg,
such as climate change, transportation, building, and energy
sustainability development director at Toyota. “As a result,
efficiency. Integrating AI into long term vision and strategy will
investors are paying more attention to sustainability. In turn
get the buy in of the employees and drive innovation at scale”
we see an increasing number of our customers passing on
these changing expectations from their shareholders into We believe that six action areas are critical as shown in Figure
strict sustainability criteria for their suppliers. Integrating 16 below.

Figure 16. Key actions for organizations to leverage to full potential of AI for climate change

Account for and take measures to

Harness AI to bring greater focus
combat the negative impact of AI
in reducing scope 3 emissions
on the climate

How
organizations Educate sustainability teams on
Collaborate with the climate can leverage AI’s how AI can make a real difference
action ecosystem full climate and educate AI teams on the
criticality of climate change
action
potential

Scale use cases on the basis of impact

Lay down the technological foundations
for your sector and emissions
for using climate-focused AI
intensity of particular functions

Source: Capgemini Research Institute analysis.

30 AI in Climate Change
Account for and take measures to combat the negative impact of AI on the climate

Despite the advances made in cognitive computing, AI Use available tools to measure, monitor and track carbon
systems can consume a lot of power and can generate high footprint of AI:
volumes of climate-changing carbon emissions. Even before
There are various approaches to assessing the environmental
beginning to deploy AI use cases, organizations need to
impact of AI. The ability to measure and report on the CO2
carefully assess the environmental impact. Organizations
footprint would help increase the general awareness of the
need to build greater awareness and take actions to ensure
severity of the problem. A working paper from researchers
that the benefits of their AI deployments outweigh their
of Stanford, Facebook AI Research, and McGill University
emissions ‘cost’.
details out a tool that could help measure both how much
A study last year found that training an AI language- electricity a machine learning project will use and how much
processing system under research stage produced 1,400 that means in carbon emissions.38 In another paper issued in
pounds of emissions – about the amount produced by flying July 2019 by the Allen Institute for AI, the authors proposed
one person roundtrip between New York and San Francisco.36 using floating point operations as the most universal and
useful energy efficiency metric.39 Another group developed a
However, it was pointed out that this 2019 study focused on
Machine Learning Calculator, which aims to quantify the GHG
natural language processing (NLP) models, which are among
emissions of machine learning.40
the largest AI models given their objective of mastering the
human language.37 As we heard from David Rolnick, chair, We tried the initial method to ascertain the GHG footprint
Climate Change AI and assistant professor of Computer of some of the popular AI use cases. The factors that were
Science at McGill University, "The training of some AI models, in used to arrive at these analyses were: duration of the process,
particular for natural language processing, can use considerable average power drained, cooling consumption factor, and unit
time and energy, though other AI models can run on a laptop." conversion factor. Our analysis shows, for instance, that the
GHG emissions produced in training and executing AI systems
vary widely from a few grams of CO2 eq. to a few kilograms,
yet they are very small in comparison with the overall GHG
footprint of a large organization (see Figure 17).

Figure 17. GHG footprint of a sample of individual AI applications

Average organization
AI use case GHG emissions produced in...
emissions

Build/training phase Run/execution phase

6 million ton of CO2

eq. – the average annual
Image recognition system for
10 kg of CO2 eq. 0.3 kg of CO2 eq. Scope 1 emissions for a
quality control at a plant
top 30 consumer products
manufacturer.

AI-based optical character 40 million ton of CO2 eq. –

recognition for a large energy 0.78 kg of CO2 eq. 0.96 kg of CO2 Scope 1 emissions for a large
company oil & gas major in Europe.

Source: Capgemini Research Institute Analysis. Carbon Disclosure Project, Climate Change 2019 reports submitted by large organizations
across sectors.

31
It is also worth noting that similar algorithms under differing Conduct an impact analysis before deploying AI at scale:
conditions (training data, server hardware, training time, Consider energy-based costs vis-à-vis benefits of deploying
cloud provider etc.) can have different GHG footprint. new AI models. Conduct a full cost benefits analysis before AI
deployment to assess the tradeoff between efficiency gains
Given this is still an emerging area of research, this
and costs (including energy costs) of deploying AI. The use
recommendation is based on what knowledge is already
of AI must be taken forward for experimentation and scale
available. In the coming years, we expect this field to develop
only if the overall impact on the climate is significantly more
much further and expect greater insights into the assessment
positive than negative.
and monitoring of AI systems.
Choose a region for your server which ensures smaller
Build capabilities to measure AI’s carbon footprint over
environmental impact: The environmental consequences
its lifecycle: AI applications result in GHG emissions due to
and the GHG footprint of AI projects can vary significantly
electricity consumption primarily during their development
on the basis of the location of the servers and data
phase which involves training the AI algorithm and the
centers hosting AI applications, due to the energy mix of a
execution phase involving running the AI application. The
particular location. Thus, organizations need to make careful
lifecycle of an AI project extends beyond these stages and
considerations about various geographies and the kind of
also involves cloud/hardware procurement, data collection,
emissions they can expect from their server or data center.
storage, among others. All these stages would have a GHG
footprint which must also be accounted for. Organizations Design and deploy efficient and sustainable AI
can also build capabilities and work with tools mentioned applications: While individual AI applications may have small
above to effectively measure, monitor and report the carbon GHG footprint in comparison to the overall GHG emissions
footprint of their AI projects. Organizations can also work of large organizations, it is important to acknowledge this
on a more detailed lifecycle assessment for AI systems in footprint and take active steps to manage it. We believe that
order to assess the environmental impacts associated with optimizing AI training and execution can significantly limit
all stages of the AI system (data collection, storage, training, its climate impact, when designed and deployed responsibly.
testing, operation, as well as the procurement of hardware We advise organizations to design efficient and sustainable
and cloud services and so on) as well as various AI-enabled AI applications, which ensures high accuracy while keeping
use cases. in mind other factors such as environmental costs, server
costs in mind as opposed to AI of the highest accuracy
and precision that has no consideration for the impact
on climate. As Lynn Kaack, chair, Climate Change AI and a
postdoctoral researcher at ETH Zurich mentioned, "When you
use technologies like IoT with AI to increase operational efficiency
in a transportation or energy system, it doesn’t mean that you’re
actually reducing emissions unless you optimize for that. If you
optimize for costs, then you don’t necessarily optimize to reduce
greenhouse gas emissions. So, it’s really important how you use
AI, and to view those technologies as a tool to help with a larger
strategy to address climate change.” Ongoing research points
to ways in which machine learning can be made greener, such
that the training of AI models is not as energy intensive as it is
slated to be. The path to making AI green depends on how it
“Tackling climate change has to be part of is designed, developed, and used.
a strategic plan for every organizations. In addition, MIT researchers have developed a new
Artificial intelligence (AI) appears naturally automated AI system for training and running certain
poised to address transformational challenges neural networks. Results indicate that, by improving the
computational efficiency of the system in some key ways,
of sustainability such as climate change, the system can cut down the pounds of carbon emissions
transportation, building, and involved – in some cases, down to low triple digits.41
Techniques such as Transfer Learning help AI practitioners
energy efficiency.”
use pre-trained models for AI algorithms that require vast
computing power (and thereby electricity), such as computer
vision and natural language processing. It significantly
reduces computing requirement for the training phase,
Meghna Tare thereby bringing down electricity consumption and GHG
Chief sustainability officer, University emissions. Also, field programmable gate arrays (FPGAs) are
of Texas at Arlington being recently adopted for accelerating the implementation
of deep learning networks due to their energy efficiency.42

32 AI in Climate Change
Organizations can focus on various training techniques to innovation, best practices, and resources. Encourage internal
design and deploy similar forms of efficient AI. collaboration across teams of AI experts, practitioners,
data scientists, that may be separated geographically, or
Collaborate with the AI expert community internal and
into various business units and functions. AI models and
external to the organization: A large community of AI
codebases with best practices for energy conservation and
experts collaborates globally across open source forums.
sustainable use must be shared internally for wider adoption
Participating in these forums helps in large scale sharing of
and overall resource saving.

Educate sustainability teams on how AI can make a real difference and educate AI
teams on the criticality of climate change

Climate change cannot be a single team’s responsibility. has a significant role to play in tackling climate change,
There has to be wider and shared accountability across many as we showcased in section 1 of this report, business and
teams and the employee base. “Climate change is everyone’s sustainability executives are not yet aligned on the idea,
responsibility,” says Zoisa Walton, CEO of Octopus Energy. neither are AI teams intentional about building sustainable AI
“Everybody is accountable for our mission and what we’re and using it to harness climate action. This is an area where
doing to support global issues like climate change. Everybody critical steps need to be taken (see Figure 18).
is responsible for how our customers react to it.” While AI

Figure 18: Building awareness and education are critical for climate action

Build awareness of sustainability teams on AI

“There’s a big component of behavioral change that needs to

be addressed when it comes to adopting advanced technology
• Educate sustainability executives and employees on AI’s
solutions such as AI for climate change because human behavior is
potential and shift leader and employee mindsets
not always 100% controllable and may influence how technology
• Bring about behaviourial change to drive greater is used.”
sustainable AI adoption
• Gabriela Prata Dias, head of the Copenhagen Centre on
Energy Efficiency.

Educate AI teams on sustainability and climate change

• Make AI teams aware of:

“The first step is to really care about climate, which means
– The climate impact of their projects and how they can
actually embedding climate impact as a priority in all that you
minimize their AI footprint and be more sustainable
do. The second thing is communicating that and empowering
– Thinking innovatively about ways in which AI can be used your employees to make climate a priority. And the third one is
to accelerate climate action for the organization and also measure – you cannot make the right decisions unless you know
learn to develop AI systems that work for sustainability what is going on both within your organization and from your
initiatives suppliers and consumers.”
• Drive top-down cultural and mindset change to bring an • Carolyn Hicks, co-founder and head of finance and
unrelenting focus on climate change, alongside an operations at Brill Power, a clean-tech startup
“AI-first” mindset.

Source: Capgemini Research Institute analysis.

33
Lay down the technological foundations for using climate-focused AI

While getting the technology foundations for AI right is Kuehne + Nagel – a global transport and logistics company
critical, only 33% have access to tools that incorporate AI based in Switzerland – launched a big data and predictive
algorithms for climate action. analytics powered platform called Sea Explorer that
allows their customers of sea freight services to track
Without this foundation, organizations will struggle to
various parameters of shipping vessels from a business
accurately track and monitor their carbon emissions and
and environment perspective.43 For instance, customers
use the resulting data to power climate-focused AI tools.
can track and optimize business KPIs such as transit times,
Organizations that are taking a lead in this area are actively
carrying capacity, and balance them with indicators of carbon
tracking thousands of data points to produce critical
emissions, allowing them to reduce carbon footprint within
insight and performance reporting. “Where possible, we
supply chain.
track our emissions using actuals from our sites.” says James
McCall, Senior Director Global Climate and Supply Chain Laying the technological foundations does not just mean
Sustainability, P&G. “This includes detailed information such as investing in the infrastructure but also bringing the right
electricity or natural gas invoices and onsite energy monitoring. talent pool into the organization for designing, developing,
This site level data, across 140 sites globally feeds into our central and running these systems. In our survey, only 32%
sustainability tracking software, resulting in over 30,000 data respondents said that they have understanding of where AI
points. This data provides the backbone for monthly/quarterly can be used and 24% said they have the required AI skills they
internal reporting, footprint reduction projects, and on an annual need to implement AI focused on the climate change space.
basis for our external citizenship report – allowing P&G to turn
data into action.”

33%
Share of organizations that have access to
tools that incorporate AI algorithms for
climate action.

34 AI in Climate Change
 Build a data-driven culture
The experts in our survey point to the absence of consider their major emissions areas and their
a data-driven culture (86%) as one of the major profile of emissions,
challenges in adopting AI for climate change. • AI teams could potentially help in identifying what
Organizations need to take a data-driven and AI systems could be harnessed to reduce this
scientific approach to tackling climate change based on the data available.
with AI. • AI and data teams also need to allow for greater
data democratization and access to trusted data
• Sustainability leaders, other than routinely
so as to allow for other teams and non-specialists
tracking and reporting emissions, need to
to benefit from it.

A data-driven culture – with a focus on climate choice. This could be particularly beneficial for AI
change – can also help attract the right talent, and teams, where talent shortages are a major problem,
help organizations differentiate as an employer of as we shared earlier in this section.

“We are very focused on collecting and accurately storing and cataloguing
emissions related data. Today, it helps us make sound decisions about
sustainability and climate change. Tomorrow, it can be useful for AI systems
that may help us with monitoring our emissions. Investing in data management
today will allow us to adopt advanced technologies such as AI for tackling climate
change in the future.”
-Maya Colambani, Director of Sustainability at L’Oréal in Brazil.

Scale use cases on the basis of impact for your sector and emissions intensity
of particular functions

Different sectors and functions in the value chain may classification tool on nearly 7,000 robots to detect
have different AI use cases that could prove to be most component failures before they happened. It was able to
beneficial. Prioritizing the right use cases for the right areas detect 72 instances of component failure that could have
is critical, as we heard from Gabriel Blanco, chair, Technology led to unplanned downtime.44 This directly results in better
Executive Committee, UNFCCC. Professor titular, Universidad industrial efficiency and thus better productivity per capita
Nacional del Centro de la Provincia de Buenos Aires, “Not CO2 emissions.
every technology or solution works in every place. It is not just a
• In the public sector, Sveaskog – Sweden’s largest forest
one size fits all. It does not happen that way. You need to start
owner and producer of sawlogs, pulpwood, and biofuel
looking at how to assess existing technologies and adapt those
– has used AI on satellite images of forests to quickly and
technologies to particular contexts and circumstances, depending
accurately identify forest areas affected by ravenous
on need and applicability.”
spruce bark beetles and to prevent them from spreading.45
• Predictive maintenance is important in the automotive This helps quick management of affected trees, and
sector. General Motors deployed a cloud-based image ultimately preserve the local ecology.

35
 Expert views on harnessing AI across functions and sectors

On food waste
“We’re not using AI right now, but we will use AI in the future, for example, to
enable people to just snap a photo of their food, rather than having to write
a description; AI would be able to help code that up. The OLIO app today is
a fraction of our full vision. We will be harnessing AI and machine learning
to continue to improve the efficiency of the platform and quality of the user
experience."
- Tessa Clarke, co-founder of Olio, a free app that connects users (households
or local businesses) who have unwanted food and those living nearby who
would like it

On power systems
“AI can be very helpful in places where information or insights are lacking or
might not be collected through traditional methods. For example, it’s very hard
to assess when the next storms or drought could affect the power systems in
Africa or Asia if you don't know power infrastructure currently operates. AI
can then help fill these data gaps with new methods. With the use of machine
vision, power infrastructure can be identified and information could be updated
regularly. Moreover, natural language processing can be used to standardize such
collected information and make it more widely accessible.”
- Johannes Friedrich, senior associate within the Global Climate Program,
World Resources Institute

On clean energy sources

"AI can provide better forecasts for electricity grids. It can forecast, for instance,
the solar power output, which helps balance supply and demand on the grid.
This can in turn result in less need for power generated with fossil fuels. AI can
also help with developing better batteries, for example, for understanding such
aspects as the degradation behavior. Or it can help to find new materials in
basic research, which is essential to develop the kind of next generation energy
technologies that are needed to transition away from fossil fuels.”
- Lynn Kaack, chair, Climate Change AI and a postdoctoral researcher
at ETH Zurich

36 AI in Climate Change
Collaborate with the climate action ecosystem

Organizations need to collaborate with a global climate advanced research, or governments and regulators to
action ecosystem, which will include peer organizations, shape policy:
partners, governments and startups. “I’d encourage businesses
• Fewer than half of organizations work with tech startups
to find others in their sector or elsewhere who want to make a
(47%) and tech companies (43%) to deploy innovative tech
positive change because there are loads of those organizations
solutions.
out there,” says Professor Tim Lenton, professor of climate
change and earth system science, University of Exeter and • Only 28% of sustainability executives said they collaborate
Director of the Global Systems Institute. “The more we see with NGOs and academia to fund high impact research and
networks of businesses and other actors coming together, the with local governments to shape policies.
better our chances.”
“The climate community across science, politics, and the private
James McCall, Senior Director Global Climate and Supply sector – including the technology community and pioneers in
Chain Sustainability, Procter and Gamble, also believes that the fields of digitalization and artificial intelligence – are not
a collaborative approach is critical. “P&G is committed that cooperating currently and are not well connected,” says Dirk
by 2030, no P&G plastic will find its way to the ocean,” he says. Messner president, German Environment Agency; Board
“But we can’t do that alone. We partner with groups like the Member, Stockholm Energy Institute. “There is a complete
Alliance to End Plastic Waste, which is a unique collaboration disconnect between the sustainability community on the one
of companies across the end to end supply chain from plastic hand and on the technological pioneers on the other hand.
producers, to manufacturing, to waste disposal and recycling. Communities are not able to benefit from technological
AEPW brings together industry, governments, and civil innovations in these fields to enhance climate protection.”
society with a common, urgent focus of ending plastic in the Organizations need to work through a cross-section of
environment." collaborations, be it with science based targets to set climate
related goals, or build cross-functional teams with academia
While organizations are focusing on industry collaborations,
and think tanks to progress and monitor climate goals and
working with tech startups or firms to find innovative tech
actions or be it working with governments to develop and
solutions is receiving less attention. At the same time, fewer
support sound policies to encourage sustainable deployment
are looking to explore newer avenues, such as academia for
of AI and other technologies to meet climate goals as well as
to develop other climate-related policies.

28%
Less than a third of organizations
collaborate with NGOs or academia for
climate action.

“P&G is committed that by 2030, no P&G plastic will find its way to the ocean. But we can’t do
that alone. We partner with groups like the Alliance to End Plastic Waste, which is a unique
collaboration of companies across the end to end supply chain from plastic producers, to
manufacturing, to waste disposal and recycling."

James McCall,
Senior Director Global Climate and Supply Chain Sustainability, Procter and Gamble

37
Harness AI to bring greater focus in reducing scope 3 emissions

Organizations need to take more active steps to reduce “saving the planet for future generations.”47 Companies are
their scope 3 emissions. If they fail to do so, reducing overall aware of how important this issue is to consumers. “It’s really
emissions will continue to be an uphill task. Organizations strategic to be committed to climate change for many reasons,”
need to work with consumers, clients and partners to said Maya Colambani, director of sustainability at L’Oréal in
increase awareness of their carbon footprint and offer low- Brazil. “One of the main reasons is that the consumer is looking
carbon solutions wherever possible. for a sustainable company. We have no way to escape from our
duty and it’s really important in order to reinforce our relationship
Organizations need to also build renewed focus on supply
with our consumer.” Speaking from the perspective of also
chains to reduce scope 3 emissions. In July 2020, the
making sustainable choices affordable to consumers, Sanand
European Commissioner for Justice committed to develop
Sule, Co-founder of Climate Connect Technologies, said, “If
legislation by 2021 that would require European companies
you give up your diesel or petrol car and if you adopt an EV as
to carry out environmental and human rights due diligence in
your next car, you take a step towards de-carbonizing the world.
their supply chains. While the details have not yet been fully
But, just because of the cost factor, consumers struggle to opt for
developed, it was explained that the legislation would be
it. I think along with consumer awareness, cost is an additional
“intersectoral, mandatory, and of course with a lot of possible
pain point.”
sanctions.”46 Speaking about the benefit of AI in supply chain,
Steve Evans director of research in industrial sustainability, Organizations can take various steps to build consumer
University of Cambridge mentioned, “Benchmarking and awareness. For example, by eco-labeling of apparels
simple information sharing in a way that’s useful is important. I organizations can create transparency about the
think that with AI, it is going to get easier to observe patterns in environmental impact of that particular apparel.
data and improve our benchmarking and thereby reduce impacts Organizations can also have dedicated consumer-facing
and risks across the supply chain.” microsites showing product makeup and origin. Moreover,
through AI-powered gamification, companies can highlight
As we heard from Kate Larsen, founder and CEO at
how consumers are bettering their carbon footprints with
SupplyESChange, an advisory network guiding environmental
every buying action and offer more eco-friendly choices
and social improvements in supply chains, “We are in early
to those who care.48 These sorts of innovations – along
days of using AI for supply chains where so many companies’
with clear guidance from standard setters – can transform
largest climate change footprints lie. AI depends on data and
awareness. “It would be nice if there was a regulatory rule that
many companies have not been tracking enough onsite verified
required less waste in our area – batteries,” says Carolyn Hicks,
supply chain GHG data yet. So, AI is not going to be impactful
co-founder and head of finance and operations Brill Power,
enough to help incentivize reducing some of the largest climate
a startup. “Then, we would not have to rely on the individual
change causing GHG emissions until we get more of the basics
decision maker’s choice - it would just happen. If battery products
done; robustly and collaboratively auditing and verifying scope 3
are properly endorsed with markings and stamps to say that a
emissions from supply chains to gain better data.”
technology is more sustainable, it makes life a lot easier not only
Our recent research study on sustainability in the consumer for consumers, but also the businesses doing the providing.”
products and retail sector found that nearly 80% of
consumers want to make a difference when it comes to

38 AI in Climate Change
Conclusion
The coming decade will be critical when it comes to climate account for and take measures to combat the negative
action and whether the world manages to avoid highly impact of AI on climate, educate sustainability teams on how
damaging climate change. Organizations across sectors AI can make a real difference and educate AI teams on the
need to take urgent actions to reduce their emissions. criticality of climate change, focus on building technological
While many traditional methods are being implemented, foundations, scale most impactful use cases, collaborate
innovation is critical, and we have found that using AI with the larger climate ecosystem and finally align scaling of
remains largely unexplored. Leveraging AI’s full climate AI use cases with emissions. Doing nothing is not an option.
action potential sustainably is mission-critical for the world
as a whole. To kick-start progress, organizations need to

39
 Research methodology
• To estimate and quantify the impact of AI on GHG • On the primary research front, we surveyed
emissions of organizations, we partnered with right. 800 executives from 400 organizations.
based on science for their expertise in XDC Model Each organization has two respondents: one
Methodology. It is the only methodology of its kind sustainability executive and one business or
to integrate a full climate model (also used by the technology executive. As well as the survey of
UN Intergovernmental Panel on Climate Change executives, we surveyed a panel of 300 experts:
(IPCC)). It is science-based, peer-reviewed, forward- regulators, academics, and AI subject matter
looking, TCFD-compatible, aligned with the EU experts.
Green Deal, transparent and Open Source (currently
• We complemented the surveys with in-depth
for academia; fully Open Source from 2021). Please
interviews of over 40 sustainability experts,
see Appendix for more details on this methodology.
business/tech experts, AI practitioners and
startups, think tanks and academicians working in
the field of AI and/or climate change.

Organization survey

Organizations by headquarters

Australia Singapore
Spain
3%

4%
Italy 4%
4% US
India 26%
4%
Brazil
5%
Sweden 5%

7% 12%
Netherlands China
7%
France 10%
10%
UK
Germany

40 AI in Climate Change
 Organizations by sector Organization by revenue

Other discrete manufacturing USD 1 billion to

less than
Automotive More than USD 50 billion
Industrial USD 5 billion
manufacturing 4% 12%
7% 26% USD 30 billion to
less than
Retail 10% USD 50 billion 9% 31%

10%
USD 20 billion to
11% less than
Process USD 30 billion
manufacturing
20% 18%
19% 23%
USD 10 billion to USD 5 billion to
Consumer products Energy and less than less than
utilities USD 20 billion USD 10 billion

Designation -Business/Technology executives

President
Senior Vice President
8%
20% Vice President
3%

Associate Director 24%

21% Director

Senior Director 25%

41
 Designation - Sustainability executives Functional Area - Sustainability executives

Senior Vice President Corporate

Climate change Social
Associate
Director 10% President Responsibility
21% 6% 26% / Corporate
Citizenship,

13% Vice 38%

President

Director 23%

28%
Senior Director Sustainability 36%

Functional area –Business executives

Human Resource
Logistics
6% Operations
Product lines 6% 32%
6%
CXO/managing director/general
management 7%

Retail operations 6%

7%
Manufacturing
8% 22%
Supply chain Corporate strategy

42 AI in Climate Change
 Expert by country of residence

Australia Singapore
Spain
3% US
Sweden 5% 15%
5% 2%
Italy
5%
India
Brazil
10%
5%

UK 10% 10%
China

10% 10%
10% France
Germany
Netherlands

Experts -by function of expertise

Regulatory professional with

Academician on AI 33% a government agency
5%
Academician on
7%
Sustainability/Environment

33% AI expert/practitioner
22%
Sustainability professional/practitioner

43
References
1. Climate.gov, Climate Change: Atmospheric Carbon Dioxide,”August 2020.
2. The Washington Post, “Earth’s carbon dioxide levels hit record high, despite coronavirus-related emissions drop,” June 2020.
3. USA Environmental Protection Agency,“Climate Change Indicators: High and Low Temperatures.”
4. National Geographic, “Sea level rise, explained.”
5. Ecowatch, “2019 Was the Oceans' Hottest Year on Record,” January 2020.
6. United Nations Office for Disaster Risk Reduction, “Human cost of disasters: an overview of the last 20 years, 2000–2019,”
October 2020.
7. IPCC, “Summary for Policymakers of IPCC Special Report on Global Warming of 1.5°C approved by governments,” October 2018.
8. BlackRock, “Our approach to sustainability,” July 2020.
9. Bloomberg, “Climate On Track to Warm at Least 3 Degrees Without Action,” December 2019.
10. Capgemini Research Institute, “The AI-powered enterprise: unlocking the potential of AI at scale,” July 2020.
11. United Nations Development Programme, Goal 13: Climate Action,”accessed September 2020.
12. Business Traveller, “Air France to use artificial intelligence to reduce CO2 emissions,” July 2020.
13. Axios, “Google drops future AI oil extraction projects,” May 2020.
14. University of Massachusetts Amherst, “Energy and Policy Considerations for Deep Learning in NLP,” 2019.
15. Guardian, “Google uses AI to cut data centre energy use by 15%,” July 2016.
16. EDF,“Methane: The other important greenhouse gas,”accessed September 2020.
17. Forbes,“How To Enhance Greenhouse Gas Monitoring From Space,”April 2019.
18. The Tribune, “Satellites reveal major new gas industry methane leaks,” June 2020.
19. BP press release,“BP North Sea deploys Mars technology in world-first methane monitoring project,”September 2019.
20. Capgemini, “Capgemini develops the mission center to monitor global CO₂ sources,” July 2020.
21. Wartsila.com,“Graciosa on the path to 100% renewable energy,”October 2018.
22. Mongabay, “‘Off the chart’: CO2 from California fires dwarf state’s fossil fuel emissions,” September 2020
23. Wall Street Journal,“California Utilities Hope Drones, AI Will Lower Risk of Future Wildfires,”September 2020.
24. Business Traveller, “Air France to use artificial intelligence to reduce CO2 emissions,” July 2020.
25. Techseen, “Alibaba’s logistics arm Cainiao to develop 1M smart vans, invest $7.3B”, May 2015.
26. Forbes,“How Ocado Is Using Machine Learning To Reduce Food Waste And Feed The Hungry,”November 2019.
27. PC Mag, “Amara’s law,” accessed October 2020.
28. Reuters, “European Parliament backs a 60% EU emissions-cutting target for 2030,” October 2020.
29. Economic Emission Intensity (EEI) = Emissions (in tons of CO2 equivalent)/GVA (in million EUR). GVA (gross value added) = EBITDA +
personnel costs. To avoid double counting of emissions, Scope 2 and 3 emissions are only counted by 50% into the total emissions.
See appendix for further explanation of the methodology.
30. IEA (2017) Energy technology perspectives 2017–catalysing energy technology transformations. https://www.iea.org/reports/
energy-technology-perspectives-2017
31. European Parliament website, “Green Deal: key to a climate-neutral and sustainable EU,” June 2020.
32. The Guardian, “EU parliament votes for 60% greenhouse gas emissions cut by 2030,” October 2020.
33. IEA, “World Energy Outlook 2020,” October 2020.
34. The World Bank, “Carbon Pricing Dashboard,” accessed October 2020.
35. Capgemini Research Institute, “The AI-Powered Enterprise”, July 2020.
36. University of Massachusetts Amherst, “Energy and Policy Considerations for Deep Learning in NLP,” 2019.
37. XD Ideas, “Yes, AI Has a Carbon Footprint – So How Do We Deal With It?” January 2020.
38. Stanford University, Facebook, Mila, McGill University, “Towards the systematic reporting of the energy and carbon footprints of
machine learning – a working paper,” February 2020.
39. Allen Institute for AI, Carnegie Mellon University, University of Washington, “Green AI,” July 2019.
40. ArXiv, “Quantifying the Carbon Emissions of Machine Learning,” November 2019.
41. Massachusetts Institute of Technology, MIT-IBM Watson AI Lab, Shanghai Jiaotong University, “Once-for-all: train one network and
specialize it for efficient deployment,” August 2019.
42. IEEE Access, “FPGA-based Accelerators of Deep Learning Networks for Learning and Classification: A Review,” January 2019.
43. Kuehne + Nagel, “Kuehne + Nagel launches Sea Explorer, the first digital platform of comprehensive sea freight service offerings,”
March 2018.
44. Iflexion, “Image Classification Everywhere in Automotive,” October 2018.

44 AI in Climate Change
45. Capgemini, “Sogeti Sweden leverage AI to hunt spruce bark beetles,” November 2019.
46. JDSPURA, “Supply Chain Managers Must Prepare for New Mandatory EU Rules on Human Rights Due Diligence,” July 2020.
47. Capgemini Research Institute, “How sustainability is fundamentally changing consumer preferences,” July 2020.
48. Martech Advisor, “How Marketers Can Use AI to Tune into Planet-Loving Prospects,” June 2017.
49. Helmke, H., Hafner, H.-P., Gebert, F. and Pankiewicz, A. (2020). Provision of Climate Services –The XDC Model. In Leal Filho, W., &
Jacob, D. (Eds.). (2020). Handbook of Climate Services. Springer. https://doi.org/10.1007/978-3-030-36875-3_12
50. O’Neill BC, Kriegler E, RiahiKet al (2014) A new scenario framework for climate change research: the concept of shared
socioeconomic pathways. Clim Change 122(3):387–400. https://doi.org/10.1007/s10584-013-0905-2
51. Smith, C. J., Forster, P. M., Allen, M., Leach, N., Millar, R. J., Passerello, G. A., and Regayre, L. A.: FAIR v1.3: A simple emissions-
based impulse response and carbon cycle model, Geosci. Model Dev., https://doi.org/10.5194/gmd-11-2273-2018, 2018; Millar,
R. J., Nicholls, Z. R., Friedlingstein, P., and Allen, M. R.: A modified impulse-response representation of the global near-surface air
temperature and atmospheric concentration response to carbon dioxide emissions, Atmos. Chem. Phys., 17, 7213-7228, https://doi.
org/10.5194/acp-17-7213-2017, 2017.
52. IEA Energy technology perspectives 2017—catalysing energy technology transformations. https://www.iea.org/reports/
energy-technology-perspectives-2017
53. Emissions include Scope 1, 2 and 3 Emissions based on the Greenhouse Gas Protocol
54. Economic Emission Intensity = Emissions (in tons of CO2 equivalent)/GVA (in million euros). GVA is gross value added = EBITDA +
personnel costs.
55. World Business Council for Sustainable Development., & World Resources Institute. (2001). The greenhouse gas protocol: A corporate
accounting and reporting standard. Geneva, Switzerland: World Business Council for Sustainable Development. Page 25.

45
Appendix
A. The XDC Model emission intensity reduction is not the same for every
sector, as they start on different levels and have different
The X-Degree Compatibility (XDC) Model, developed by right. potentials.
based on science (right.), calculates the contribution of a
company, portfolio or any other economic entity to climate The XDC Gap: is the difference between the Sector
change, answering the question: How much global warming Target XDC and a chosen other XDC (e.g. Baseline XDC or
could we expect, if the entire world operated at the same Scenario XDC). It expresses the alignment with the <2°C
economic emission intensity as the entity in question under target – the larger the XDC Gap, the further away the
a given scenario? Results are expressed in a tangible degree company is from achieving <2°C-alignment. A negative
Celsius (°C) number: the XDC. This science-based temperature (≤0) XDC Gap signifies, that the company is already
alignment metric offers unprecedented transparency to operating in alignment to the <2°C target. The XDC Gap
companies, banks, investors, and the public on climate risk enables the comparison of temperature alignment of
and climate alpha. Thus enabling the transition to a <2°C companies or other economic entities from different
economy and to fulfilling the Paris Climate Agreement (Paris sectors, as both need to achieve different Sector Target
Alignment). The XDC Model is the only methodology of XDCs and thus have a different goal to achieve.
its kind to integrate a full climate model (also used by the
UN Intergovernmental Panel on Climate Change (IPCC)). It II. Portfolio XDC Metrics
is science-based, peer-reviewed, forward-looking, TCFD-
compatible, aligned with the EU Green Deal, transparent and These metrics can be aggregated to a portfolio of
Open Source (currently for academia; fully Open Source from companies to allow for a comparison on group-level. The
2021).49 aggregation of metrics results in a Portfolio Baseline
XDC, a Portfolio Target XDC and the corresponding
I. The XDC Climate Metrics
Portfolio XDC Gap. It is essential to note that only the
Portfolio XDC Gaps may be compared, as the variation in
The XDC Model generates four basic climate metrics: the
the sector-makeup within different portfolios results in
Baseline XDC, the Scenario XDC, the Sector Target XDC,
different Portfolio Target XDCs.
and the XDC Gap.
To compare the temperature alignment of Climate AI
Champions to the rest of the sample, we have analyzed
The Baseline XDC: expresses the °C of global warming
the individual companies in our survey and consolidated
if the entire world economy were to operate with the
them on a group – or “portfolio” – level according to the
same Economic Emission Intensity (EEI) as the company
classification above. These aggregated metrics result in:
in question (see below for calculation process). The
underlying assumption is that the historic ratio between • A Portfolio Baseline XDC – XDC metric of the portfolio of
emission and GVA growth remains stable. These growth companies
trends have been defined by SSP 2 of the set of Shared
Socioeconomic Pathways (SSPs).50 • A Portfolio Target XDC – XDC target to be aligned with the
Paris Agreement
The Scenario XDC: is calculated with the same approach • The corresponding Portfolio XDC Gaps – the difference
as the Baseline XDC, but with different assumptions on between the above two metrics, signifying the gap that
the growth of GVA and emissions. These growth rates can remains to be closed if the organization is to meet the Paris
be defined individually, for example based on a climate Agreement.
strategy, other Shared Socioeconomic Pathways, or any
type of scenario. This approach has been used in chapter We have further broken down the resulting Portfolio XDC
1 to calculate the XDCs and corresponding climate Gaps based on Scope 1 and 2 emissions (see Appendix D on
pathways for different sectors, under the assumption “Emissions Accounting” for more details on emission scopes).
of emission reduction rates through AI that have been The Portfolio XDC Gaps make it possible to understand
determined in the survey. the alignment of the corresponding groups to the Paris
Agreement and to assess the effect of mature and aligned
The Sector Target XDC: expresses the temperature that AI and climate strategies. A similar analysis is possible for
a company must achieve, to be aligned to a given global scope 3 emissions. However, we have limited our analysis to
warming scenario. This target is specific to the company’s scope 1 and 2, owing to several challenges associated with
sector. This is rooted in the idea that the feasibility of

46 AI in Climate Change
completeness, accuracy, and adequacy of scope 3 emissions • The calculation of global warming associated with an
data. amount of GHG emissions and other climate pollutants is
based on the FaIR (Finite Amplitude Impulse Response)51
The XDC Gap can only be closed by decoupling the increase
climate model and consists of the following steps:
of emissions from economic growth, resulting in a reduction
of Economic Emission Intensity (EEI), for example through the – Quantification of induced increase of atmospheric
implementation of technology. This analysis demonstrates greenhouse gas concentration
the potential of strong AI capabilities in alignment with
– Radiative Forcing associated with increase of
a mature climate strategy to achieve a reduction in EEI,
concentration
indicated by a closing XDC Gap.
– Global Warming (= XDC)

III. Calculation of the Baseline XDC

IV. Calculation of the Sector Target XDC
The calculation process of the XDC Model is divided into
The calculation of the Sector Target XDC is based on
three major sections: (1) Economic Emission Intensity of
three steps: (1) calculating the median EEI for each
a company, (2) Projection of cumulative global emissions
sector, (2) calculating the sector-specific EEI target
from base year until 2050, and (3) Calculation of the
pathway, (3) scaling the EEI pathway to the global GVA
global warming resulting from the emissions determined
and (4) calculating the global warming resulting from the
in (2).
emissions determined in (3).
• Economic Emission Intensity. In a first step, the Economic In step (2), the GVA growth rate follows the SSP2
Emission Intensity (EEI) for the company in question scenario and the emission reduction rates are calculated
ascertained. EEI is defined as the amount of greenhouse from the IEA emission budgets
gas emissions (in tons CO2e) that the company emits per
• Emission budgets. The International Energy Agency (IEA)
1 million euro gross value added (GVA). GVA is defined as
has developed emissions budgets for all sectors in a 2
the sum of EBITDA and Personnel Costs. The emissions are
Degree Scenario (D2S) and a Beyond 2 Degree Scenario
entered into the model as Scope 1, 2 and 3 Emissions as
(B2DS).52 This is rooted in the idea that the feasibility of
defined by the Greenhouse Gas Protocol (see appendix e
emission intensity reduction is not the same for every
for further information). To avoid double counting, Scope
sector, as they start on different levels and have different
1 emissions are fully counted (100%) while Scope 2 and 3
potentials.
emissions are only partially counted (50%).
• EEI Target Pathways. The sector-specific emission budgets
• Global Emissions. The EEI from step one is projected
are used to compute EEI pathways for each sector until
until 2050 and scaled to a global level to compute the
2050. Based on these pathways, we compute the global
global emissions from the base year until 2050. This is the
emissions from the base year until 2050.
quantity of emissions that would reach the atmosphere
if the entire world economy were to operate as emission • Global Warming. Just as with the calculation of the
intensively as the economic entity in question. Baseline XDC, the emissions that have been computed in
the previous step are fed into the full climate model to
• Global Warming. In a third step, a full climate model (the
compute the global emissions from the base year until
FaIR model, see below) is used to calculate the amount
2050. The result is the Sector Target XDC.
of global warming that would occur if the amount of
emissions calculated in step two were to be released into In the EEI Pathway analysis in chapter 1, the EEI Target
the atmosphere. By doing this, insight can be provided pathway (which is the basis for the calculation of the Sector
on the level of global warming that would result if every Target XDC) has been used.
company were to operate as emissions-intensively as the
company under consideration.

47
B. Modeling the impact of AI adoption overall GHG emission reduction rates for the time span
2026 to 2030.
on GHG emission reduction
• The overall GHG emission reduction rates were converted
To understand the contribution that AI can make to climate
into yearly reduction rates.
action, we modeled the impact of AI on the sector-specific
Economic Emission Intensity (EEI) along with our partner • We assumed that the GHG emission reductions as
right. based on science. The EEI of a sector is calculated estimated in the survey uniformly apply to all emission
as the amount of its emissions53 (measured in tons of CO2 scopes.
equivalent) per unit gross value added (the sum of EBITDA
• We also assume an optimistic future in the next ten years,
and personnel costs).54
where organizations scale AI at double the level that exists
EEI Pathways (which are part of the XDC Model as previously today and realize up to 50% of GHG reduction through
explained in appendix a) allow us to assess to which extent a AI-enabled use cases as reported in our survey. It is to take
sector is able to decouple the growth of GHG emissions form into account the losses that invariably occur when an AI use
the growth of economic activity (as measured by gross value case is scaled to the organizational level. For instance, if a
added). For each of the five sectors, we computed the EEI use case is expected to reduce emissions by X% individually
Pathway for the “AI-enabled” scenario in two steps. For the (in one project/location/facility), when scaled organization-
base year 2018, the median EEI from the sector is used. Then, wide, the resulting benefit for the entire organization will
standard GVA growth rates and GHG emission reduction rates be less than X% – owing to various losses and additional
estimated from our survey were applied to calculate the EEI complexities in the process. This allows us to explore the
for the following years. highest potential benefits of AI-enabled use cases for
climate action.
We calculated the reduction rates of the GHG emissions for
the time span 2018 to 2030 as follows. These individually by Capgemini defined “AI-enabled” EEI
pathways are then compared with right.’s standard EEI
• The overall GHG emission reduction rates given for the
Baseline pathway and the EEI Target pathway which has
time span 2017 to 2019 (Figure 4) and the time span 2020
been calculated in line with the B2DS scenario as explained in
to 2025 (Figure 6) were linearly extrapolated to obtain
Appendix A.

Economic Emission Intensity Models for various sectors

Evolution of Economic Emissions Intensity Pathways for the utilities sector

5500.00
(tCO2e/Mio EUR Gross Value Added)
Economic Emissions Intensity

5000.00

4500.00

4000.00
14%
3500.00

3000.00

2500.00
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 42% 49%

B2DS pathway (for limiting global warming to 1.750C per IEA and Paris Agreement)
“AI-enabled” pathway
Baseline (business-as-usual) pathway

48 AI in Climate Change
 Evolution of Economic Emissions Intensity Pathways for the consumer retail sector
1150.00
1050.00
950.00
850.00
Economic Emissions Intensity (tCO2e/Mio EUR Gross Value Added)

750.00 17%
650.00
550.00
38% 45%
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

Evolution of Economic Emissions Intensity Pathways for the wholesale retail sector
3050.00

2550.00
4%
2050.00

1550.00
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 38% 45%

Evolution of Economic Emissions Intensity Pathways for the oil and gas sector
5500.00
5000.00
4500.00
4000.00 7%
3500.00
3000.00
2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 36% 43%
B2DS pathway (for limiting global warming to 1.75 C per IEA and Paris Agreement)
0

“AI-enabled” pathway
Baseline (business-as-usual) pathway

Disclaimer from right. based on science regarding the XDC Model

right. based on science GmbH is not an investment adviser and makes no representation regarding the advisability of investing in any particular
company, investment fund or other vehicle. A decision to invest in any such investment fund or other entity should not be made based on any of
the statements set forth in this publication. While right. based on science GmbH has obtained information believed to be reliable, right. based
on science GmbH shall not be liable for any claims or losses of any nature in connection with information contained in this document, including
but not limited to, lost profits or punitive or consequential damages.
The information used to compile this report has been collected from a number of sources, especially from FactSet Research Systems and
Urgentem. Some of its content may be proprietary and belong to right. based on science GmbH.
The information contained in this research report does not constitute an offer to sell securities or the solicitation of an offer to buy, or
recommendation for investment in any securities within any jurisdiction. The information is not intended as financial advice. This research
report provides general information only. The information and opinions constitute a judgment as at the date indicated and are subject to
change without notice. The information may therefore not be accurate or current. The information and opinions contained in this report have
been compiled or derived from sources believed to be reliable and in good faith, but no representation or warranty, express or implied, is made
by right. based on science GmbH as to their accuracy, completeness or correctness and finally right. based on science GmbH does not warrant
that the information is up-to-date.
The contents, works and information published in this document are subject to German copyright. Any kind of duplication, processing,
distribution, storage and any kind of exploitation outside the limits of copyright requires the prior written consent of the respective copyright
holder. If you want to use the information provided, please contact us.
© right. based on science GmbH

49
C. Complete list of AI use cases for climate action
Use cases marked in green are the most adopted use cases by help analyze historical reasons of failures and combine it with
industry. real time data such as images from production line to predict
upcoming failures of machines. In one such example, an
Automotive sector:
automotive giant managed to increase throughput by 25–30
Working example: Machine stoppage can lead to wastage vehicles per day. This prevented unplanned stoppages of
of work-in-process inventory and loss of productivity. AI can assembly line leading to overall improvement in productivity,
and more efficient utilization of power and resources.

Emission scope
Part of
AI for climate action use case improved by the
value chain
use case
i. Using AI to improving engine efficiency and reduce use of fuel 3
ii. Using AI to optimize performance based on driving conditions while increasing
3
efficiencies
iii. New fuel (hybrid) systems using machine learning 3
Research & iv. Designing lighter components for overall vehicle weight reduction 3
Development
v. Power output optimization for electric vehicles based on driving conditions and
2
tackling range.
vi. Virtual testing and simulation for new model/product testing and design 1
vii. Setting of internal carbon pricing for all components, supply chain and production
1
through AI-based insights
i. Raw material supply chain tracing using machine learning/AI 3

ii. Tracing carbon emission of supply chain using big data analytics/AI 3

iii. Logistics optimization using AI 3

Supply chain &
manufacturing iv. Predictive maintenance at manufacturing facility to increase efficiency and lower
1
usage of resources
v. Sustainable prototyping and testing processes e.g., use of additive manufacturing
1
in prototyping and recycling waste produced
vi. Improving energy efficiency of manufacturing facilities 1

i. Tracking use phase carbon footprint of vehicles using AI 3

ii. Collecting and keeping track of end of life vehicle using machine learning 3
Sales &
aftersales
iii. Using ML/AI in selecting and retrofitting components 1

iv. Predictive maintenance of vehicles to ensure compliance 3

i. Ride sharing portal for day to day employee commute using blockchain/AI 1

ii. IT operations optimization using edge computing/deep learning 1

Internal iii. Using AI to enable employees to work remotely through advanced remote
1
operations working technologies, and reducing travel needs
iv. AI to assess and mitigate financial and insurance risk due to climate change 1

v. Automation of carbon emissions, tracking and reporting 1

50 AI in Climate Change
Manufacturing sector:
Working examples: properties and dimensions of the product, thus reducing
• Prototyping can be resource-intensive based on the wastage.
complexity of desired outcome and properties of the • In AI-enabled predictive maintenance, computer vision can
product being designed. AI can help design prototypes help identify and track early signs of failure or inefficiency
more efficiently while meeting or exceeding desired in manufacturing assets. A timely fix flagged in time can
help avoid wastage

Emission scope
Value chain AI for climate action use case improved by the
use case
i. Improving machine/asset efficiency of end products to reduce use of fuel/electricity 2

ii. Designing lighter components for reduction in product weight 3

iii. Energy consumption measuring platform using AI 2

iv. Setting of internal carbon pricing for all components, supply chain and production
Research & 1
through AI-based insights
Development
v. Selection of low-carbon raw material and packaging components through AI/big
1
data machine learning
vi. Optimizing low-carbon processes and operations (such as carbon capture and
1
operational reuse) through AI control systems
vii. Adoption of green hydrogen substitutes for industrial operations through AI
1
assisted operations.
i. Raw material supply chain tracing using machine learning/AI 3

ii. Tracing carbon emission of supply chain using big data analytics/AI 3

iii. Logistics optimization using AI 1

Supply chain &
iv. Predictive maintenance at manufacturing facility to increase their efficiency
manufacturing 1
resulting in less use of resources
v. Sustainable prototyping and testing processes e.g., use of artificial intelligence in
1
prototyping and recycling of waste produced in testing
vi. Electrification of heating and industrial processes and machinery utilizing AI
2
supported grid balancing and energy storage
i. Recollecting and keeping track of end of life products/by-products using ML/AI 3
Sales &
ii. Using ML/AI in selecting and retrofitting components that are recollected 3
aftersales
iii. Shifting to product-as-a-service model supported by AI based utilization prediction 3

i. Ride sharing portal for day to day employee commute using blockchain/AI 1
ii. Energy consumption optimization of IT infrastructure, employee work facilities using
1
edge computing/deep learning
Internal iii. Using AI to enable employees to work remotely through advanced remote working
operations 1
technologies, and reducing travel needs
iv. AI to assess and mitigate financial and insurance risk due to climate change 1

v. Automation of carbon emissions, tracking, and reporting 1

51
Energy and Utilities sector:
Working example: Utilities need to accurately forecast AI can help utilities process huge data coming from historical
energy demand to avoid energy waste. An accurate forecast consumer usage, weather conditions, monthly/annual
also allows grid operators to more effectively use renewable trends, monthly attrition/churn rate/volume growth to more
energy in combination to the traditional sources of energy. accurately predict energy usage and maximize renewable
energy production.

Emission scope
Value chain AI for climate action use case improved by the
use case
i. Energy consumption measuring platform using AI/ML 1
Research & ii. Renewable energy modeling/analysis/prediction using ML/AI 1
Development
iii. Setting of internal carbon pricing for all components, supply chain and production
1
through AI-based insights
i. Raw material supply chain tracing 3
ii. Algorithms to automatically identify defects and predict failures without
1
interrupting operations (predictive maintenance)
iii. Carbon-capture technologies assisted with AI production 1

iv. Pipeline maintenance and leakage detection 1

v. Forecasting using machine learning, data mining about weather, environment, grid
1
operations.
Production/
vi. Tracing carbon emission of supply chain 3
Generation
vii. Logistics optimization 3

viii. Use of smart grid 1

ix. Use smart energy meter 1

x. Minimizing the production of waste methane instead of flaring through AI/machine
1
learning production optimization
xi. Capture and management of by-products (such as methane) from well production
1
utilizing AI prediction and anatomy detection. E.g., vapor/venting recovery units
i. Consumer-behavior analysis using advanced analytics 3

ii. Making electric supplies more reliable and bespoke to consumer needs 1
Sales &
aftersales iii. Generating and building AI-based consumer insights and awareness on energy
3
utilization, home efficiencies and carbon footprint
iv. Incentivizing electrification of building heating through incentives for utilizing
2
renewable, energy-enabling predictive grid balancing.
i. Ride sharing portal for day to day employee commute using blockchain/AI 1
ii. Energy consumption optimization of IT infrastructure, employee work facilities
1
using edge computing/deep learning
Internal iii. Using AI to enable employees to work remotely through advanced remote working
operations 1
technologies, and reducing travel needs
iv. AI to assess and mitigate financial and insurance risk due to climate change 1

v. Automation of carbon emissions, tracking, and reporting 1

52 AI in Climate Change
Consumer products and retail sector:
types, external variables (route disruption, weather
Working examples: In supply chain, AI can help process conditions, etc.) to arrive at the shortest, fuel efficient routes
multiple variables – sources, destinations, routes, product while fulfilling business constraints.

Emission scope
Value chain AI for climate action use case improved by the
use case
i. Improving efficiency of product development and personalization using AI 1
ii. Using AI/machine learning to find and utilizing biodegradable substitutes and
1
reducing packaging footprint
Research & iii. Selection of low-carbon raw material and packaging components through AI/big
1
Development data machine learning
iv. Setting of internal carbon pricing for all components, supply chain and production
1
through AI-based insights
v. Reducing waste 3

i. Automate assembly and identification of parts to reduce manual intervention 1

ii. Predictive maintenance of machines 1

iii. Using AI to optimize changes in products to achieve operational efficiency 1

Supply chain &
manufacturing iv. Tracing carbon emissions within the supply chain using big data analytics 3
v. Enhanced traceability of raw materials to optimize production using machine
1
learning/AI
vi. Improving energy efficiency of manufacturing facilities 2

i. Effective inventory prediction/optimization and management 1

ii. Route optimization using advanced analytics 1

iii. Autonomous vehicles within distribution centers 1

Sales &
aftersales
iv. Cold chain logistics optimization 3

v. Last-mile delivery and electric vehicles operations optimization 3

vi. Bridging the gap between supply and demand using predictive analytics 3

i. Ride sharing portal for day-to-day employee commute using blockchain/AI 1

ii. Energy consumption optimization of IT infrastructure, employee work facilities
1
using edge computing/deep learning
Internal iii. Using AI to enable employees to work remotely through advanced remote working
operations 1
technologies, and reducing travel needs
iv. AI to assess and mitigate financial and insurance risk due to climate change 1

v. Automation of carbon emissions, tracking, and reporting 1

53
D. Emissions accounting
The Greenhouse Gas Protocol is a commonly used standard of purchased electricity consumed by the company.
for the accounting of GHG emissions. It defines three scopes Purchased electricity is defined as electricity that is
of emissions: 55 purchased or otherwise brought into the organizational
boundary of the company. Scope 2 emissions physically
• Scope 1: Direct GHG emissions. Direct GHG emissions occur
occur at the facility where electricity is generated.
from sources that are owned or controlled by the company,
for example, emissions from combustion in owned or • Scope 3: Other indirect GHG emissions scope 3 is an
controlled boilers, furnaces, vehicles, etc.; emissions optional reporting category that allows for the treatment
from chemical production in owned or controlled process of all other indirect emissions. Scope 3 emissions are a
equipment. consequence of the activities of the company but occur
from sources not owned or controlled by the company.
• Scope 2: Electricity indirect GHG emissions scope
2 accounts for GHG emissions from the generation

Figure: GHG emission scopes distribution

CO2 CH4 N2O HFCs PFCs SF6

Scope 2 Scope 1
INDIRECT DIRECT

Scope 3 Scope 3
INDIRECT INDIRECT

purchased goods
and services transportation
and distribution
leased assets investments
capital goods company
facilities
franchises
processing of
fuel and energy employee sold products
related activities commuting
company leased assets
business tavel vehicles
transportation use of sold
and distribution waste generated products
in operations end-of-life treatment
of sold products

Upstream activities Reporting company Downstream activities

Source: Greenhouse gas protocol, “Technical Guidance for Calculating Scope 3 Emissions v1,” accessed October 2020.

54 AI in Climate Change
55
About the Authors
Anne-Laure Thieullent Vincent de Montalivet
Vice President, Artificial Intelligence & Analytics Head of Sustainable AI, Capgemini
Group Offer Leader [email protected]
[email protected] Vincent leads Sustainable AI, part of Perform AI
Anne-Laure leads Perform AI, Capgemini’s Artificial –Capgemini Group AI & Analytics offering. He has
Intelligence & Analytics group offer. She works with been working in sustainable ICT for more than ten
clients to scale Artificial Intelligence so it infuses years and has held IT project management and
everything they do, and they become AI-driven, business development roles across consulting,
data-centric and innovative. With 20 years of SMEs, local authorities and NGOs.
experience in big data, analytics and AI systems,
from design to production roll-out, her passion is
to bring companies what they need to transform
themselves into intelligent enterprises. AnneLaure
is committed to guiding clients to increase
activating their data, while cultivating the values of
trust, privacy and fairness. This is the foundation
on which technology, business transformation
and governance come together, to leverage the
positive business outcomes of AI.

Dr. James Robey Kiri Trier

Global Head of Environmental Sustainability, Director, Capgemini Invent Germany
Capgemini [email protected]
[email protected] Kiri Trier is the Sustainability Lead (director) for DACH
Dr James Robey has led the sustainability agenda at Capgemini Invent. She advises Capgemini clients on
at Capgemini since 2008, creating and driving a the transformation of their traditional business into
broad ranging program to reduce the Group’s a sustainable business by defining and implementing
own environmental impacts whilst identifying the right sustainability strategy,integrating all
opportunities to support Capgemini’s clients with sustainability requirements like ESG, SDGs and GRI.
their own sustainability challenges. In addition, he She has over 15 years of experience in innovation
teaches at a number of leading universities on the management and business modelling.
topic of Sustainable Business.

Florent Andrillon Jean-Baptiste Perrin

Vice President, Energy Transition Global Leader, Vice President, Invent for Society Global Leader,
Capgemini Invent Capgemini Invent
[email protected] [email protected]
Florent Andrillon is the energy transition lead for Jean-Baptiste leads Capgemini Invent’s social
Capgemini Invent, where he advises companies from impact business initiative : Invent for Society. With a
different sectors on their energy and low carbon strong background in public sector and sustainable
transformation projects. development, he is a Vice-President within Citizen
Services at Capgemini Invent France and currently
develops business synergies with our most recent
acquisitions including social impact agency Purpose.

Michiel Boreel Philippe Vié

Global Chief Technology Officer at Sogeti and Vice President, Capgemini group Energy, Utilities
Executive Vice President of Capgemini and Chemicals sector leader
[email protected] [email protected]
Michiel Boreel is the Global Chief Technology Philippe has over 25 years of Energy and Utilities
Officer at Sogeti and Executive Vice President of industry experience and dedication, with a strong
Capgemini.Since 1994 Michiel leads a small think tank focus on Utilities transformation projects, including
within SogetiLabs called VINT (Vision, Inspiration, sector and digital technologies levers. Philippe
Navigation Trends) with the goal of inspiring leads the Capgemini landmark publication, the
customers to apply new technology in business. World Energy Market Observatory (WEMO).
Constant theme in all of the SogetiLabs research is
the question what impact IT has on society and how
technology can be applied to create meaningful
business innovation. Since the start of the institute it
has published over 15 books and many reports.

56 AI in Climate Change
 Katja van Beaumont Marc Rietra
Vice President, Content lead - Consumer Products, Vice President, Content lead - Consumer Products,
Retail, Distribution, Capgemini The Netherlands Retail, Distribution, Capgemini The Netherlands
[email protected] [email protected]
Katja is a thought-leader at Capgemini in the Marc is a thought-leader at Capgemini in the
private industries. With over 20 years of experience private industries. With over 20 years of experience
in delivering transformational business initiatives in delivering transformational initiatives he has
enabled by technology, she is a visionary leader, been leading customers to success, always bringing
elevating aspirations and bringing about a lasting in a unique vision ensuring meaning and long-term
positive impact for all stakeholders. value creation.

Steve Jones Anne-Laure Cadene

Chief Data Architect, Capgemini Group Portfolio Vice President, Capgemini Digital Engineering and
[email protected] Engineering Services
Steve Jones is the Chief Data Architect for [email protected]
Capgemini’s Group Portfolio, with over a decade in Anne-Laure is a Research & Innovation director at
Big Data and Cloud he’s been at the front of helping Digital Engineering and Engineering Services of
our clients adopt new technologies to drive new Capgemini. She has been with Capgemini since
business value and become more data driven. Steve January 2019 and is based in France. She has over
was one of the first people to propose using mobile 20 years of experience, after a PhD in
applications and analytics to help track and manage fluid mechanics.
individuals carbon footprint and has continued to
be at the front of Capgemini’s CSR and Diversity
initiatives.

Jerome Buvat Amol Khadikar

Global Head of Research and Head of Capgemini Senior Manager, Capgemini Research Institute
Research Institute [email protected]
[email protected] Amol is a senior manager at the Capgemini
Jerome is head of the Capgemini Research Research Institute. He leads research projects
Institute. He works closely with industry leaders on key frontiers such as artificial intelligence,
and academics to help organizations understand sustainability and the future of work to help clients
the nature and impact of devise and implement data-driven strategies.
digital disruption.

Amrita Sengupta Yashwardhan Khemka

Manager, Capgemini Research Institute Manager, Capgemini Research Institute
[email protected] [email protected]
Amrita is a manager at the Capgemini Research Yash is a manager at the Capgemini Research
Institute. She tracks the patterns of digital Institute. He likes to follow disruption fueled by
disruptions across industries and their impact technology across sectors.
on businesses.

Shahul Nath
Consultant, Capgemini Research Institute
[email protected]
Shahul is a consultant at the Capgemini Research
Institute. He keenly follows disruptive technologies
and its impact on industries and society.

The authors would like to especially thank Subrahmanyam KVJ and Ankita Fanje from the Capgemini Research Institute and Hannah Helmke,
Sarah Söding, Liv Hammann, and Sebastian Müller from the startup right. based on science for their contribution to this research. In addition,
the authors would like to thank, Shobha Meera, Maitreyee Mathur, Barkha Gosain, Charlie Weibel, Bobby Ngai, Sara Moubarak, Nisheeth
Srivastava, Pierre-Denis Autric, Jérôme Fenyo, Gunnar Menzel, Kees Jacobs, Ashvin Parmar, Narendra Bhosekar, Arthur Arrighi De Casanova,
Philipp Harker, Valerie Perhirin, Moez Draief, Marianne Boust, Martin Bennetzen, Padmashree Shagrithaya, Dan A Simion, Dr. Jingyuan Zhao,
Fabian Schladitz, Ivar Aune, Raul Bartolome, Robert Engels, Simon Gosset, Martin Chauvin, Jordan Toh, Anne-Violaine Monnie-Agazzi and
Soumik Das for their contribution to this research.

About the Capgemini Research Institute

The Capgemini Research Institute is Capgemini’s in-house think tank on all things digital. The Institute publishes research on the impact of
digital technologies on large traditional businesses. The team draws on the worldwide network of Capgemini experts and works closely
with academic and technology partners. The Institute has dedicated research centers in India, Singapore, the United Kingdom, and the
United States. It was recently ranked number one in the world for the quality of its research by independent analysts.
Visit us at www.capgemini.com/researchinstitute/

57
Leverage Data & AI’s
transformative power
with Capgemini
Perform AI
Activate data. Augment intelligence. Amplify outcomes.
Artificial Intelligence (AI) has crossed the threshold of pilots, and entered the wider market. 53% of organizations have managed
to scale AI projects in production – but only 13% overall have rolled out multiple AI applications across numerous teams, as per
the findings of the Capgemini Research Institute report – The AI Powered Enterprise: Unlocking the potential of AI at scale.
Perform AI, Capgemini’s comprehensive portfolio of Data, Analytics, Intelligent Automation and Artificial Intelligence, provides
your organization with the ability to leverage the full transformative power of Data & AI at scale. By activating data and insights
at the heart of your business, in your everyday decisions and actions, you will augment your organization’s intelligence, and
amplify business outcomes you expect from Data & AI.

Data & AI for trusted business outcomes

Leading organizations are today already leveraging the transformative power of Data and AI to design and launch new intelligent
products, services, business models – even creating new markets. They are also finding efficiencies, optimizing processes and
reducing costs, becoming more agile so they can grasp new opportunities, and developing greater resilience to sudden crises, via
Intelligent Business, IT & Security operations.
Our teams have a proven track record of delivering tangible business outcomes globally and at scale, across industries
and sectors:

Faster Better
Increased customer 2X customer 24
sales 25% insights Impact service Languages

Organizations successfully scaling AI A global consumer products firm One of the world’s largest biotech
have seen more than a 25% increase successfully engaged with 1 billion companies used our award-winning*
in sales of their products and services. people using customer data & insights virtual assistants to automate
with double the impact in half the outbound telephone calls in
time for half the cost. 24 languages.

Operational 40% Cost Fraud 10X

excellence cheaper savings €100m detection ROI

A leading producer of packaging and A major international retailer used A European government agency has
paper used our AI solution to reduce our powered solution to optimize delivered a 10x return on investment
its cost of handling invoice processing sales forecasting which directly drove thanks to our AI-powered fraud
queries by 40% overall. €100m of inventory costs savings. detection systems

*Capgemini wins 2020 Artificial Intelligence Breakthrough award for best virtual agent solution

58 AI in Climate Change
The right team to scale Data & AI
With 25,000 Data & AI at scale practitioners, working for more than 800 clients worldwide, supported by AI Centers of
Excellence in all regions, Capgemini’s capabilities are unmatched.
We show organizations not only how transformative Data & AI could be, we actually deliver it at the heart of the business. We
define the right strategy with industrialization, operationalization and scale in mind. We ensure you take the right approach to
finding actionable, trusted insights for your teams, as well as including them in the transformation journey. And we build the
right solutions for your business and implement them with the right continuous deployment and operations to maximize their
positive impact.

Do good with Data & AI

Because we are convinced about the transformative power of Data & AI to build positive futures for humans and society, we
partner with you to leverage Data & AI in an ethics by design and human-centered way. We work with your teams to develop the
right data and AI-powered leadership, mindset, culture and ways of working that fit your organization’s values. Find out more in
our report AI and the Ethical Conundrum: How organizations can build ethically robust AI systems and gain trust.
We help you balance operational excellence with business innovation to be resilient to current and future crises, to win in your
market and protect your workforce. And we help you find opportunities to proactively use Data & AI for their positive impact on
your business and our society.
"We help you shift gears, scale trusted Data & AI solutions for the best business outcomes, and Get The Future You Want."

Anne-Laure Thieullent
Vice President, Artificial Intelligence &
Analytics / Perform AI Group Offer Leader

Sustainable AI
Achieve low-carbon AI and power your sustainability goals

What’s the connection between AI and sustainability?

Both are probably at the top of your strategic agenda. But have you thought about how they impact one another?
AI will transform global productivity and efficiency, equality and inclusion, and a host of other areas critical to the
Paris Agreement pledges. There is much to be gained from considering how your AI project can improve not just your
organization’s prospects but all our futures.
Sustainable AI solutions can reduce pollution, shrink your carbon footprint and waste, and help counter global warming.

Our approach
We are already helping organizations harness AI to achieve their sustainability goals. This means thinking about AI from
two perspectives:
• Building Green AI by implementing the most resource efficient solutions: Achieving this involves understanding
and mitigating the energy and carbon footprint of machine learning and its infrastructure.
• AI for climate action: Efficiency gains are usually visible at the bottom line, but they can also support your
sustainability vision and the UN’s Sustainable Development Goals. Taking leadership here can be a powerful indicator
of an organization’s values.

59
Capgemini Perform AI
Accelerated time to value for business outcomes with
Trusted Data & AI at scale

Energy
Manuf, Auto & Telco Media Financial Services
CP & Retail Public Sector Utilities &
Life Sciences Tech & Insurance
Chemicals
Augmented Infusing AI to Augmented Deep Customer Sustainability Right Product,
operations better serve operations Engagement and powered by Right Price
and immersive citizens and for operational Operational Data & AI and
customer governments efficiency & risk Excellence Operational
engagement reduction Excellence

Data & AI for

Customer Intelligent Enterprise Data Partnerships

First Industry Management & Data disruptors

Intelligent Applications for the Intelligent & Resilient Enterprise

Strategy Build, Deploy, Manage, Operate

AI, Analytics Data Intelligent Process

AI Activate
Science & 890 Automation

Transformation Target Scalable, Trusted AI and Intelligent Automation

and foundations for Augmented BI & Data Viz for human & machine
execution of AI @Scale solutions collaboration

AI Reimagine AI & Data Engineering

Disrupt, Invent and Hybrid Cloud Technology

Implement “the next” platforms for Trusted Data &
business, “AI First” AI @Scale transformation

Privacy Trust & Ethics Security Sustainability

60 AI in Climate Change
Capgemini Invent
A preferred partner to help you solve trust challenges using AI and
data – for climate action

Value proposition and approach

As a globally renowned technology and digital leader, Capgemini inherits the responsibility, the ambition, and the
means to contribute to solving major societal questions that shape our world – and at Capgemini Invent we are
contributing to making this ambition a reality. Invent for Society showcases how social impact is part of the fabric
of what we do for our clients every day. For more information, please visit our Invent for society page and have a
look to our dedicated “environment” pillar presentation on YouTube:

Why us ?
We partner with clients to:
• Increase the share of low-carbon fuels for all types of transportation – including electric, natural gas, and
hydrogen
• Help industrial organizations, buildings and cities to reduce their energy consumption and CO2 footprint
• Leverage AI and digital to reduce CO2 emissions and create new business models

Want to know more about what we can do for you ?

Please contact

Valerie Perhirin Paul Dixon

Managing Director in charge of Vice-President,Public Sector
Insight-Driven-Enterprise / Sustainability
[email protected] [email protected]

Jean-Baptiste Perrin Florent Andrillon

Vice-President, Invent for Society Vice-President, Energy &
Global Leader Utilities / Sustainability
[email protected] [email protected]

61
For more information,
please contact:
Global

Anne-Laure Thieullent Vincent de Montalivet

Vice President, Artificial Intelligence and Analytics Group Head of Sustainable AI, Capgemini
Offer Leader [email protected]
[email protected]

Valerie Perhirin Florent Andrillon

Managing Director in charge of Insight-Driven-Enterprise, Vice President, Energy Transition Global Leader,
Capgemini Invent Capgemini Invent
[email protected] [email protected]

Arthur Arrighi De Casanova

Director, Sustainability & Energy Transition,
Capgemini Invent
[email protected]

APAC France
Dr. Jingyuan Zhao Marion Gardais
AI Leader VP Presales & Solutioning
[email protected] [email protected]

Germany Europe
Fabian Schladitz Robert Engels
AI Centre of Excellence Leader AI Leader
[email protected] [email protected]

India North America

Padmashree Shagrithaya Dan A Simion
AI Centre of Excellence Leader AI Centre of Excellence Leader
[email protected] [email protected]

United Kingdom
Joanne Peplow
AI Centre of Excellence Leader
[email protected]

62 AI in Climate Change
Discover more about our
recent research

The Automotive Sustainability in CPR Sustainability in energy

Industry in the Era of and utilities
Sustainability

AI and the Artificial Intelligence in

The AI-powered ethical conundrum Manufacturing
enterprise: Unlocking the Operations
potential of AI at scale

The Art of customer centric Why addressing ethical

Accelerating Automotive’s Artificial Intelligence questions in AI will
AI Transformation: How benefit organizations
driving AI enterprise-wide Raising the bar in customer
can turbo-charge experience
organization value

63
 Conversations Towards Emotional Intelligence Turning AI into concrete
Ethical AI: Views from The essential skillset for value: the successful
Experts and the age of AI implementers’ toolkit
Practitioners

Building the Retail Superstar Agile at scale:

Voice on the Go How can How unleashing AI across Four recommendations
auto manufacturers functions offers a multi-billion to gain enterprise-wide
provide a superior in-car dollar opportunity agility
voice experience

Scaling innovation

64 AI in Climate Change
Subscribe to the latest research
from the Capgemini Research
Institute
Receive advance copies of our reports by scanning the QR code or visiting
https://www.capgemini.com/capgemini-research-institute-subscription/

65
66 AI in Climate Change
67
 About
Capgemini
Capgemini is a global leader in consulting, digital transformation,
technology and engineering services. The Group is at the forefront
of innovation to address the entire breadth of clients’ opportunities
in the evolving world of cloud, digital and platforms. Building on
its strong 50-year+ heritage and deep industry-specific expertise,
Capgemini enables organizations to realize their business ambitions
through an array of services from strategy to operations. Capgemini
is driven by the conviction that the business value of technology
comes from and through people. Today, it is a multicultural
company of 270,000 team members in almost 50 countries. With
Altran, the Group reported 2019 combined revenues of €17 billion.

Visit us at

www.capgemini.com

MACS_15072020_SD

The information contained in this document is proprietary. ©2020 Capgemini.

All rights reserved.