In collaboration with MEXT Technology Center

Unlocking Value from

Artificial Intelligence
in Manufacturing
WHITE PAPER
DECEMBER 2022
 Cover: Jian Fan, Getty Images – Inside: Getty Images

Contents
3 Foreword

4 Executive summary

5 Introduction

6 1 Unlocking value in manufacturing through AI

8 2 Shedding light on common barriers to industrial AI adoption

11 3 A collection of AI applications in manufacturing

17 4 A step-by-step approach to implementing scalable industrial AI applications

21 Conclusion

22 Contributors

25 Endnotes

Disclaimer
This document is published by the World
Economic Forum as a contribution to a
project, insight area or interaction. The
findings, interpretations and conclusions
expressed herein are a result of a
collaborative process facilitated and
endorsed by the World Economic Forum
but whose results do not necessarily
represent the views of the World Economic
Forum, nor the entirety of its Members,
Partners or other stakeholders.

© 2022 World Economic Forum. All rights

reserved. No part of this publication may
be reproduced or transmitted in any form
or by any means, including photocopying
and recording, or by any information
storage and retrieval system.

Unlocking Value from Artificial Intelligence in Manufacturing 2

December 2022 Unlocking Value from
Artificial Intelligence
in Manufacturing

Foreword

Özgür Burak Akkol Jeremy Jurgens

Chairman, Turkish Employers’ Managing Director,
Association of Metal Industries World Economic Forum

Türkiye has established itself as a key global player Ministry of Industry and Technology and established
in advanced manufacturing and aims to boost by the Turkish Employers’ Association of Metal
its position through Fourth Industrial Revolution Industries (MESS) – joined the World Economic
technologies. In recent decades, the country has Forum’s Centre for the Fourth Industrial Revolution
made significant efforts to position itself as a global Network, the foremost platform helping leaders
innovation hub, excelling in developing state-of-the- anticipate emerging technologies and drive their
art technologies in ground-breaking companies in inclusive and sustainable adoption. The network links
various fields. Artificial intelligence (AI) technology on-the-ground experience and action with global
applications are part of this effort. In principle, AI network-based collaboration, learning and scaling.
could unlock more than $13 trillion in the global
economy and boost GDP by 2% per year.1 This white paper is an output of the ongoing
However, companies struggle to tap into the value partnership between the Forum’s Platform for
that AI applications can create. Shaping the Future of Advanced Manufacturing and
Value Chains and Platform for Shaping the Future
This paper seeks to uncover the hidden potential of of Technology Governance: Artificial Intelligence
AI in the manufacturing sector and the respective and Machine Learning, the Centre for the Fourth
end-to-end systems by providing practical use Industrial Revolution Affiliate in Türkiye and MESS.
cases and critical enablers to help harness its It highlights case studies from organizations
potential. Coupled with the energy crisis and on the impact, feasibility and scalability of AI in
material shortages facing the world, manufacturing manufacturing. It identifies several opportunities and
players need to go beyond traditional operating lessons from the community on how to increase
methods to drive efficiency and sustainability. operational efficiency, sustainability and workforce
engagement in manufacturing and value chains by
The twin challenges of technological progress using AI.
and socio-political distress call for new forms
of cooperation that respond to heightened We hope this report will provide decision-makers
demand for localization while recognizing the with a better understanding of how to unlock the
drivers of connectivity that shape global impact. untapped potential of industrial artificial intelligence
Acknowledging this, the Centre for the Fourth (AI). We look forward to collaborating with you to
Industrial Revolution in Türkiye – mandated by the deploy these technologies responsibly.

Unlocking Value from Artificial Intelligence in Manufacturing 3

Executive summary
Recent global developments and an ever-growing The consultations show that leading manufacturers
list of shocks and disruptions have put further have successfully overcome the challenges
strain on already shaken global value chains. mentioned above, implementing a variety of AI
The complexity of current challenges impacting applications and achieving a positive impact
manufacturing and value chains calls for the need on operational performance, sustainability and
to go beyond the traditional means of driving workforce engagement, mainly in six areas: health
productivity to uncover the next wave of value for and safety, quality, maintenance, production
businesses, the workforce and the environment. processes, the supply chain, and energy
Artificial intelligence (AI) is a crucial enabler of management.
industry transformation, opening new ways to
address business problems and unlock innovation While opportunities enabled by AI in manufacturing
while driving operational performance, sustainability are promising and attracting many leaders,
and inclusion. Even though the impact of AI organizations are looking for a common framework
applications on manufacturing processes is known, that outlines how to implement AI solutions and
the full opportunity from their deployment is still to ensure a successful return on investment.
be uncovered due to a number of organizational
and technical roadblocks. Based on the consultations, this white paper
presents one step-by-step process as an example
Recognizing this need, the Centre for the Fourth of how it is possible to overcome barriers, using
Industrial Revolution Türkiye, together with the the AI Navigator2 developed by the INC Invention
World Economic Forum’s Platform for Shaping Center as a reference:
the Future of Advanced Manufacturing and Value
Chains and Platform for Shaping the Future of Phase 0: Initiation to build the fundamentals –
Technology Governance: Artificial Intelligence and strategy, data and workforce
Machine Learning, convened industry, technology
and academic experts to shed light on these Phase 1: Ideation to identify potential use cases and
challenges and propose a step-by-step approach conduct a pre-selection
to overcome them. The consultations revealed six
main challenges hindering the adoption and scaling Phase 2: Assessment to select use cases and
of AI applications in manufacturing: identify priorities via gap analysis

1. A mismatch between AI capabilities and Phase 3: Feasibility to complete all required tests
operational needs and studies

2. The absence of a strategic approach and Phase 4: Implementation, which requires iteration
leadership communication and piloting using agile project management

3. Insufficient skills at the intersection of AI and Moving forward, the World Economic Forum and
operations the Centre for the Fourth Industrial Revolution
Türkiye will continue to work closely with
4. Data availability and the absence of a data stakeholders in the Centre for the Fourth Industrial
governance structure Revolution Network and across industries to
accelerate the journey to capture value from AI
5. A lack of explainable AI models in manufacturing in manufacturing globally. It will offer the Turkish
Employers’ Association of Metal Industries (MESS)
6. Significant customization efforts across Technology Centre as a unique testing and
manufacturing use cases collaboration system for businesses to pilot new
AI applications and foster a collaborative approach
among a diverse group of stakeholders to ensure
the right AI capabilities are built in manufacturing
and rolled out worldwide.

Unlocking Value from Artificial Intelligence in Manufacturing 4

Introduction
Companies across value chains are now facing While manufacturers acknowledge the importance
an energy crisis and material and key component and urgency of embedding AI in their processes
shortages, even as they are still recovering from and and while leading companies have already
adapting to COVID-19 impacts. The complexity of internalized it in their business processes, many
the challenges impacting operations calls for the are becoming disillusioned with their efforts to
need to go beyond the traditional means of driving capture value from it and lag in developing the right
productivity to uncover the next wave of value and AI capabilities.
address sustainability and workforce challenges.
Artificial intelligence (AI) can enable a new era in the Understanding the purpose and role of AI is key to
digital transformation journey, offering tremendous solving manufacturing challenges. With a problem-
potential to transform industries to gain greater oriented approach, AI efforts can be linked to
efficiency, sustainability and workforce engagement clear business targets, giving business units and
by generating new insights from large amounts of business functions a joint interest in making the
data. However, despite this promising value creation transformation successful.4
potential, the deployment of AI in manufacturing
and value chains is still below expected levels. This white paper sheds light on the benefits that can
be achieved through industrial AI and the successful
Based on a global survey conducted over the last AI applications implemented across industries,
four years of more than 3,000 companies across lessons learned and tangible impacts. Consultations
industries and geographies, a growing number of conducted with the multistakeholder initiative
companies recognize the business imperative to community find that industrial AI helps people work
improve their AI competencies: in a smarter, safer and more efficient way. However,
to unlock its full potential, companies require an
– 70% of respondents understand how AI can understanding of current barriers to adoption and a
generate business value structured approach to overcome them. Therefore,
this paper also presents one example of a step-by-
– 59% have an AI strategy in place step guide to successfully implementing scalable
industrial AI use cases.
– 57% affirm that their companies are piloting or
deploying AI.

Despite these trends, only 1 in 10 companies

believe they generate significant financial benefits
with AI.3

Unlocking Value from Artificial Intelligence in Manufacturing 5

1 Unlocking value
in manufacturing
through AI
AI applications in manufacturing help
increase operational performance,
drive the sustainability agenda and
empower the workforce.

The artificial intelligence (AI) revolution allows the Türkiye, find that AI can help drive a step-change in
conversion of large amounts of data into actionable manufacturing, yielding significant benefits in three
insights and predictions that can provide impetus to categories (figure 1):
data-driven processes. Manufacturing companies
capture value from AI using different mechanisms, – Operational performance by automating
the most common being eliminating redundant and optimizing routine processes and tasks,
work, solving existing problems and revealing increasing productivity and operational
hidden value by analysing and recognizing patterns efficiencies, improving quality (e.g. reducing
in data. AI is applied to augment tasks such as defects, forecasting unwanted failures) and
classification, continuous estimation, clustering, optimizing production parameters
optimization, anomaly detection, rankings,
recommendations and data generation to solve – Sustainability by optimizing material and
industrial problems.5 energy usage, increasing energy efficiencies,
reducing scrap rates and extending machine
Consultations with senior executives from the World lifespans
Economic Forum’s Platform for Shaping the Future
of Advanced Manufacturing and Value Chains and – Workforce augmentation by guiding the
Platform for Shaping the Future of Technology decision-making process and parameter
Governance: Artificial Intelligence and Machine setting, enhancing the accuracy of predictions
Learning, as well as members and partners of and forecasting, reducing repetitive tasks and
the Centre for the Fourth Industrial Revolution increasing human-robot interactions

Unlocking Value from Artificial Intelligence in Manufacturing 6

FIGURE 1 Dimensions of value creation with AI in manufacturing

Operational performance

Performance (e.g. yield optimization)

Throughput (e.g. fewer unwanted breakdowns,

decreased lead time)

Quality (e.g. fewer process defects and failure rates)

Business uptime (e.g. productive time and capacity)

Workforce augmentation

Decision-making and planning support

Collaboration

Prediction and forecasting accuracy

Task automation

Risk (e.g. feedback mechanism to avoid incidents

and alarms)

Sustainability

Material efficiency

Energy efficiency (e.g. energy savings

and thermal efficiency)

Machine lifetime

Scrap rate and used material

Unlocking Value from Artificial Intelligence in Manufacturing 7

 2 Shedding light on
common barriers to
industrial AI adoption
Implementing AI solutions requires
continuous project management
efforts, expectation management and
the necessary resources.

Despite this potential, companies have not yet fully components, understanding them will help identify a
realized the vision of AI-powered manufacturing pathway to implement scalable AI applications.
systems. To unlock the untapped value of industrial
AI, pinpointing the source of a company’s struggles Consultations with the community of over 35 senior
and defining the roadblocks open a new path operations executives, technology experts and
to think through and derive the right solutions to academics have identified six challenges hindering
overcome them. As the barriers to AI adoption stem the adoption of AI in manufacturing and value
mainly from organizational, strategic and technical chains (figure 2).

FIGURE 2 Barriers to AI adoption in manufacturing

Mismatch between AI Absence of a strategic Insufficient skills at the

capabilities and operational approach and leadership intersection of AI and
needs communication operations

Data availability and Lack of explainable AI Significant customization

absence of a data models in manufacturing efforts across
governance structure manufacturing use cases

Unlocking Value from Artificial Intelligence in Manufacturing 8

Mismatch between AI capabilities and operational needs

Manufacturers have often selected AI projects operations. This causes a mismatch of expectations
based on existing technical capabilities instead of and hinders their wider adoption in manufacturing.
focusing on the impact on business operations. Building a solid business case with a problem-
The match between business pain points and AI oriented approach that clearly defines business
technologies is not always thoroughly considered. needs and evaluating the value of an AI solution
Therefore, AI solutions may be technically feasible compared to alternative solutions are the first steps
but fail to solve a relevant, impactful problem in in overcoming that barrier to adoption and scale.

Absence of a strategic approach and leadership communication

A clear company-wide AI strategy and communication to workforce reluctance. As AI is changing the ways
plan are often ignored. Without the right sponsors of working, communicating the strategic approach,
and committed leaders to start the dialogue and benefits and new processes can help increase end-
collect the buy-in from end-users, the onboarding of users’ willingness to embrace it in their routines.
AI applications across the company can’t occur due

Insufficient skills at the intersection of AI and operations

External consultants or information technology and implementation by cross-functional teams

(IT) experts who have a limited understanding with diverse expertise at the convergence of
of the manufacturing requirements on the shop IT, operational technology (OT), data and AI
floor often lead AI projects. However, to be technologies. This requires upskilling the workforce
successful, AI applications require development and attracting new talent in manufacturing.

Unlocking Value from Artificial Intelligence in Manufacturing 9

Data availability and the absence of a data governance structure

Applying machine learning models requires training Creating a single source of information ensures
on large amounts of data to recognize patterns and that businesses operate based on standardized,
relationships.6 However, manufacturing companies relevant data across the organization. To overcome
often rely on small data sets and fragmented data, this challenge, sharing data across companies’
hindering the accuracy of the resulting insights. boundaries can support joint efforts to adopt
Even when available, these data sets may not artificial intelligence techniques in the manufacturing
represent appropriate failure cases or relevant sector and rely, in turn, on a set of organizational
process situations and are mostly not interoperable. and technological success factors.7

Lack of explainable AI models in manufacturing

The perception of AI models as complex, non- and accurate and have a warning mechanism in
transparent and uninterpretable systems hinders place to minimize risks. Explainable AI tools and
their deployment. Manufacturers need AI models techniques allow experts to obtain justifications for
that are either open and transparent to build trust in their results in a format that manufacturing users
the predictions and specific results or interpretable can understand. The greater the confidence in the
for domain experts to accept them. AI-provided AI-powered output, the faster and more widely AI
predictions need to be meaningful, explainable deployment can happen.

Significant customization efforts across manufacturing use cases

Factories are complex engineered systems and AI testing of AI models still need manual intervention
models need configuration to be adapted to each for customization, which is not yet fully automated.
process and conform to its constraints. Hence, it Additionally, industrial companies struggle to find
is not possible to simply apply trained AI models commercially available hardware and software
or pipelines from one manufacturing use case with off-the-shelf AI features that require minor
to another. The design of the machine learning customization.
pipeline and the pre-processing, training and

Shedding light on these challenges and understanding them can help identify the right solutions and
approaches to overcome them.

Unlocking Value from Artificial Intelligence in Manufacturing 10

3 A collection of AI
applications in
manufacturing
AI applications can boost operational
performance and lead to a positive
impact on sustainability and workforce
engagement.

Consultations with over 35 senior operations To illustrate the potential and feasibility of AI in
executives and technology experts find that leading manufacturing, the creation of an industrial AI
manufacturing companies have successfully use case library with input from the community
managed to approach and overcome the has started. The 23 use cases collected across
challenges mentioned above by starting with their different industries cover six main application areas:
business needs, outlining a clear strategy, building health and safety, quality, maintenance, production
cross-functional capabilities and putting a stronger process, supply chains, and energy management
focus on data governance, and selecting AI models (figure 3).
that meet their needs.

They have implemented a variety of AI applications

that have boosted their operational performance
and led to a positive impact on sustainability and
workforce engagement.

Unlocking Value from Artificial Intelligence in Manufacturing 11

 FIGURE 3 Leading manufacturers are implementing a variety of AI applications

1
Health and safety
– Employee health & safety:

AI in
incident prevention
– Process safety:
2 advanced alarm analytics

manufacturing
use cases
Quality
– Quality inspection in assembly
– Quality assurance/
3 defect inspection
– Quality testing
– Quality prediction

Maintenance
4
– Machine health monitoring:
predictive maintenance
5 – Maintenance planning
Source: Company interviews

Energy management Supply chains Production process

– Energy optimization – Future demand and price – Process optimization
forecasting
– Electricity demand forecasting – Line balancing
– Supply chain control tower
– Heating and cooling optimization – Product design and development
– Warranty and service
– Process parameter optimization
management
– Production planning/
decision support

The use cases collected provide valuable insights investments is usually tangible within 1-2 years.
indicating the business need, the solution After piloting the AI applications in one division,
implemented and the impact achieved. The manufacturing companies either have already
applications show that the return on investment deployed to multiple divisions or have the vision
(ROI) is positive and the payback period of the to scale.

Unlocking Value from Artificial Intelligence in Manufacturing 12

 TA B L E 1 A collection of AI in manufacturing use cases

Use case Company Sector AI application Impact

Model designed as an
experienced operator/engineer – Total time of alarm floods
in continuous estimation and decreased by 40%
classification of alarms, detection
– Number of alarms
of nuisance alarms, alarm flood
Process safety: decreased by 50%
Tüpraş, analysis and recommendation
advanced Energy – Time efficiency: Alarm
Türkiye of better configurations. Root
alarm analytics rationalization meetings
causes, next-best actions and
set points extracted from the shortened from 4 hours
Health & safety

historical data through basic to 30 minutes

descriptive analytics and data
science pre-process techniques

Image recognition by monitoring – Unsafe situations and

the shop floor with existing actions reduced by 70-
Employee
cameras, receiving real-time 80%
health & Intenseye,
Manufacturing alert notifications and enhancing
safety: incident USA – With a safer environment,
employee health and safety
prevention a more productive
(EHS) to eliminate life-altering
workforce with increased
injuries
business uptime created

– Up to 40% savings
Examining the effective achieved in energy use
parameters on the frames – Scrap rate reduced while
Real-time spot Martur being welded in robotic spot
ensuring sustainability in
weld quality Fompak, Automotive weld stations (weld quality)
production
prediction Türkiye and predicting the spot nugget
diameter realized in line in real – Costs reduced by 60%
time by preventing the use of
excess welding materials

Visual inspection to ensure – Productivity increased

the coating quality is good by by 11%
Detection of
Bosch, checking parts and searching for – 15 million parts checked
carbon coating Automotive
Türkiye coating defects in four different had no incidents
defects
classes: scratches, damages,
black in black, silver
Quality

Optimizing quality inspection

of customized products by
Quality deploying cloud services and
– Productivity increased by
assurance a federated learning approach
Huawei, 30-40%
with federated Production (local data collected, global
China – Lead time reduced
learning in optimum interpolated and in turn
control shared back to all local facilities
without disclosing sensible
product or process data)

Explainable computer vision

methods used to support factory – 10x less implementation
Quality workers in detecting assembly effort expended
inspection Ethon AI, errors on printed circuit boards – Trustworthiness of the
Electronics
in assembly Switzerland (e.g. missing, faulty, or wrong system increased with
verification components) via a human-AI the explainable model
interface (camera system with
live feedback)

Unlocking Value from Artificial Intelligence in Manufacturing 13

 Use case Company Sector AI application Impact

– Report preparation time

Visual inspection of fibre ratio in for customer complaints
yarn content using microscopic and analysis expected to
Karsu,
Quality testing Textile images to check production
Türkiye decrease by 90%
quality and to analyse customer
– Expert requirement
complaints
for the subject will be
eliminated

– Reduction of false-reject
Visual inspection of the quality rate by an average of 88%
Quality of pharmaceuticals while AI – Detection rate increased
Körber
inspection recognizes patterns instead of by an average of 38%
Digital, Pharmaceuticals
in drug-and measuring physical image values, – Approximately 2x faster
Germany
patient safety which decreases the false-reject time-to-market achieved
of products
Quality

(transferability) in vision
setup

An AI engine that predicts the – Machine capacity

Schneider demagnetization voltage to increased
Predictive
Electric, Electronics reduce the number of iterations – Capex investment reduced
quality
France during relay tests in residual – Rejections reduced
current device product range

Through combination of digital – Productivity and quality

twin and innovative AI, process sustainability increased
Obeikan anomaly conditions and drivers
– Overall equipment
Quality Digital detected
Chemicals effectiveness in PET lines
prediction Solutions,
improved by 20%
Saudi Arabia Statistical process control
algorithm, a proven approach of – Customer complaints
quality control, used reduced

Providing automated software to

take preventive actions early in the – Alloy use reduced by 9%
Process Fero Labs, production process with explainable at steel mills
Steel
optimization USA AI models to reduce raw material – Failure rate eliminated
use and minimize costs and
emissions during steel production

– Productivity increased
AI-based video analytics to label by 25%
Production process

the actions of manual tasks

– By increasing quality and
Khenda, to eliminate operator-related
Line balancing Automotive efficiency, error costs
Türkiye errors and improve manual
manufacturing processes and eliminated and waste
optimize line balancing and defective products
avoided

Generating insights into the

complex interactions between – Defects reduced to 0%
hundreds of process parameters from a 6% of historical
Production and their impact on final quality defect rate
Dataprophet,
parameter Foundry by using deep learning algorithms
South Africa – Number of quality stops
optimization
reduced from 81 to 20
Application then prescribes next-
per week
best step to optimize production
without poor quality

Unlocking Value from Artificial Intelligence in Manufacturing 14

 Use case Company Sector AI application Impact

– Service call rate improved

by 15.3%.
– 17.8% of test capacity
increased by decreasing
Improving cooling test
Advanced the test time from 80
performance in different and
decision min. to
dynamically changing climatic
support Arçelik, – 65 min.
Home Appliance conditions to shorten the test
system on Türkiye – 17.8% in energy savings
duration by an in-house decision-
performance
Production process

making system based on AI and per unit in LPT (long

test
machine learning (ML) performance test) system
16.7% of warranty cost
improvement per unit
achieved

Implementing AI-enabled process

controls to manage catalyst – Overall batch cycle time
Process ingestion based on pressure reduced by 22% and
GEP, USA Chemicals
management and temperature changes in need to add new reactor
the reactor and to manage the
capacity alleviated
transfer rates

– Downtime reduced by
Predictive Detecting the machinery 90%
maintenance: failure mode by collecting the – Maintenance cost
machine health Sensemore, continuous vibration data through reduced by 25%
Cement
monitoring and Türkiye fault estimation, early warning – Machine life increased
maintenance and maintenance planning with by 20%
planning AI on fans and electric motors – Operation productivity
improved by 25%

Predicting the concentration

of contaminative particles
before it can negatively impact
– 70% reduction in the
Maintenance

The Center the production yield, allowing

costs of unplanned
for Intelligent chamber cleaning to be
Predictive downtime
Maintenance Electronics performed in a proactive manner
maintenance – Competitiveness
Systems Solution is for dry etching
(IMS), USA chamber in semiconductor improved
manufacturing to monitor the
deposit accumulation process
inside the processing chamber

AI-based predictive solutions

for industrial robots for an
– 50% reduction in
automotive manufacturing
unplanned downtime
Predictive Predictronics, client to monitor welding robot
Automotive – Inefficient maintenance
maintenance USA health and ultimately predict
and prevent failure events and practices reduced
schedule maintenance, saving
time, money and resources

Unlocking Value from Artificial Intelligence in Manufacturing 15

 Use case Company Sector AI application Impact

Forecasting future demand for

products, services and raw – Forecasting accuracies
Future demand
SmartOpt, material prices by automatically for the next 6 months
and price Chemical
Türkiye training AI model and tuning the reached 85-99%
forecasting
model parameters automatically
without user input
Supply chain

Component-based prediction – Prediction accuracy

granularity as the components of increased from 70%
Warranty
Tofaş, the vehicles affected by different (manual prediction) to
and service Automotive
Türkiye factors to dwetermine warranty 95%, which resulted in
management
expenditure of the coming years
reduction of reserve fund
for the sold vehicles
by 10% per year

– 5.09% gain in thermal

Analysing thermal efficiencies, efficiency, which
ambient conditions to control translates to 9M lbs/yr
real-time natural gas optimization and energy cost savings
Energy Canvass AI, Food and
and consumption targets across achieved
optimization Canada Beverage
turbines; electricity production
– CO2 emissions reduced
targets and steam demand from
– Overall equipment
downstream boilers
effectiveness improved

– Prediction accuracy
Predicting electricity usage score varies between
with changing periods in the 12
Energy management

80-95% for the total

Electricity different regions of the factory consumption of the
Ford Otosan,
demand Automotive with a model that worked
Türkiye factory
forecasting automatically in predetermined
– The pre-emptive right
periods and eliminated error
factors to buy the electricity
is reserved with the
electricity provider

– Improved the energy

management system’s
Optimizing energy management energy efficiency in
system (vehicle temperature) of electric vehicles by
electric vehicles by simulating the 10% on average, with
Heating Makinarocks, control environment using a deep a maximum increase of
and cooling Republic of Electric vehicle neural network-based dynamics 25%
optimization Korea model and implementing a – An additional 5-7%
reinforcement learning method increase in driving
to improve energy efficiency distance per every 10-
through optimized control inputs
15% increase in energy
management system
energy efficiency

Source: Company interviews

Unlocking Value from Artificial Intelligence in Manufacturing 16

4 A step-by-step
approach to
implementing
scalable industrial
AI applications
It is possible to uncover AI’s untapped
potential with a holistic approach.

Digital and AI can power a new era for continuous step approach and the required toolset, the
improvement in manufacturing beyond the traditional manufacturing sector can gain new perspectives to
means of driving productivity, thereby unlocking overcome the most pressing challenges.
additional value. Although leading manufacturers
have already captured significant benefits from AI To do so, the INC Invention Center has developed
applications, some are still trying to get started and the AI Navigator,8 a structured toolset to help
are looking for a common framework that paves the companies reach the untapped potential of AI and
way for the deployment of AI in manufacturing with identify use cases with the strongest chance of
a positive return on investment. successful implementation. The step-by-step guide
presented here (figure 4) – based on the AI Navigator
This study has demonstrated that it is possible and consultations from the initiative community –
to uncover AI’s untapped potential with a provides an example of how to develop application-
holistic approach. The development of AI-based specific roadmaps and companywide actions, from
applications starts by laying the groundwork idea generation through evaluation and feasibility
and some fundamental steps. With a step-by- analysis to successful implementation.

Unlocking Value from Artificial Intelligence in Manufacturing 17

 FIGURE 4 A step-by-step guide to implementing AI in manufacturing

Phase 0

Initiation
Kick-off preparation

Strategy Data Workforce

Strategy and leadership commitment Data governance structure Team setting & capability building

Phase 1 Phase 2 Phase 3 Phase 4

Ideation Assessment Feasibility Implementation

Problem – solution fit AI readiness level Implementation concept Product-market fit
check

– Discover – Infrastructure – Business feasibility – Agile approach

– Define – Data – Technical feasibility – Explainable model
– Empathize – Process – Data testing – Testing
– Confront – Mindset/culture – Technology scouting – Evaluation
– Competence validation – Deployment

Use case pre-selection Use case prioritization Use case decision Use case validation

Source: Adapted from the work of the INC Invention Center

Phase 0: Initiation

AI is an emerging discipline intended to create a 2. Data governance will be an indispensable

system that amplifies and expands human abilities. part of the process. While collecting the data,
To implement a holistic approach independent from it is necessary to build a hub for data flows to
industry specifics, there are three fundamentals: manage data availability, usability, integrity and
strategy, data and workforce. security. AI applications are only built when the
data is available; therefore, the most prominent,
1. Strategic approach and leadership easily applicable and transferable data structure
commitment to cultivate the AI transition needs to be built beforehand.
comprehensively, from business units to
production facilities. To successfully translate 3. Cross-functional teams with multi-disciplinary
the strategy into action, leading manufacturers in-depth knowledge in IT, OT, data, analytics and
have placed significant efforts on change technology engineering and how to organize
management by actively communicating the set these with business acumen. Therefore, it
vision and its benefits, investing in workforce is crucial to build an agile team structure
upskilling and infrastructure upgrades and for AI projects that revives the collaboration
establishing a digital company culture. For AI between the technology team and business
governance and data ethics mechanisms to be professionals. AI solutions require collaboration
responsible and explainable, they have to be with different skill sets and expertise. Even
incorporated into this strategy. though external sources are used in the
transition, in-house experts need to be upskilled.

Unlocking Value from Artificial Intelligence in Manufacturing 18

Phase 1: Identification

AI projects have the greatest probability of success define the main pain points and thus develop the
when businesses conduct a critical early analysis relevant requirements for the solutions needs to
of potential use cases resulting from a problem- be involved.
oriented approach. After setting the stage in Phase
0, this first phase of the AI Navigator enables Close communication and collaboration with
the early prioritization of use cases based on a relevant colleagues and end-users at an early stage
problem-solution-fit analysis. help ensure the developed solution meets their
needs and allows them to adapt progressively to
To discover a set of potential value-adding use the introduced change. Ultimately, this will increase
cases, a cross-functional team that can analyse and the acceptance of the deployed application’s use.9

Phase 2: Assessment

The second phase of the AI Navigator methodology be formulated. Later implementation can only be
focuses on a much more precise assessment made possible after all to-dos have been processed
of the maturity level of the company-specific and the resulting gap between the maturity
AI use case preselected in phase 1 in order to levels has been closed. Consequently, based on
establish the current status quo.10 This phase the assessment results, a statement about the
aims to determine whether the use case can be expected implementation effort can be made with a
successfully integrated with the existing data set, high degree of accuracy.
process perspective, infrastructure and culture/
mindset. Subsequently, a rough solution hypothesis The completion of phase 2 further narrows down or
is to be defined, whereby the target maturity can focuses on the use cases that offer comparatively
be specified. Usually, there is a gap between the high added value and fulfil the necessary
current and target maturity levels. To close this gap, prerequisites for later implementation.
well-founded recommendations for action must

Phase 3: Feasibility

The third phase of the AI Navigator focuses on available on the market or at the research and
business and technical feasibility. In addition to ROI development stage are analysed, considering
estimations, critical elements of this phase include: the company’s individual tech stack. In the best
case, existing solutions can be used directly or
– Data testing: This ensures that enough at least built upon.
structured and labelled data sets of the right
quality are available to produce the required – Competence analysis: If the company requires
results. This provides initial insights into the or desires in-house development, a well-
general feasibility of the use case in preparation founded competence analysis of the employees’
for implementation. This is essential to the skills is necessary to decide which cooperation
development of the specific technology partners are needed externally and which skills
concept.11 can be covered or built up internally.12

– Technology scouting: To check whether the During this phase, the solution hypothesis
previously established solution hypothesis might becomes increasingly concrete. If the resulting
be developed or purchased, it is recommended implementation concept deviates from the original
to carry out the technology scouting as a solution hypothesis, it may be necessary to perform
next step. In technology scouting, solutions the individual steps in phase 3 again.

Phase 4: Implementation

The focus of the fourth phase is to clearly define the the process, ensure the implemented application
implementation roadmap. Industrial AI applications addresses the need of the end-user, and provide a
need modification, testing and validation of models space for innovation and co-creation.
with iterations, which takes time. Using agile
project management methodologies across an It is not enough to pilot an AI use case to leverage
open, collaborative environment, including internal the potential of existing data. Scaling the solutions
and external team members, can help streamline developed is critical to success. For this, it is also

Unlocking Value from Artificial Intelligence in Manufacturing 19

 necessary to consider the data available and the The systematic approach such as that of the AI
parameters of related use cases, data access, data Navigator enables an additional benefit in terms of
governance and security, and the required skills the comparability of different ideas and use cases
early on.13 and alignment of the strategy. While it is often
possible to develop many technology-oriented
Not many organizations have experience with ideas, the challenge lies in prioritizing different
AI. It is crucial to ensure employees join the topics, even in the same department. Using an
transformation to a data-driven future, eliminate approach like the AI Navigator and a well-defined
fears and prejudices, and establish a culture open to evaluation logic allows for the comparison of
failure.14 To build and sustain trust in AI, transparent different use cases across the four phases to
and explainable AI models must be implemented determine how relevant they are and what the
and the domain expert must be aligned with AI potential implementation effort will be in terms
recommendations. Leading the projects with agile of technology, mindset and culture. In addition
sprints makes this progress smoother. The maturity to building cross-functional teams and involving
of AI models improves over time, which means it is the domain expert in the development process,
not necessary to change the model according to the the cultural component of a guide such as the
first sprint results, as adding additional data sources AI Navigator enables the early identification of
and increasing the size of available data yields potential action points.16 Developing the AI literacy
better results. To accelerate the model training curve of domain experts by collaborating with academia
collaboratively, “federated learning (an approach and capability-building centres without focusing on
to machine learning in which the training data are vendor-specific tools is crucial to building the team.
not managed centrally)”15 can distribute the effort
for multiple parties across multiple decentralized
edge devices or servers holding local data samples
without exchanging them.

CASE STUDY

Pilot
implementation at
MEXT Technology
Center

The Turkish Employers’ Association of Metal Industries solution to serve as a testbed for new technologies,
(MESS), one of Türkiye’s largest and most active employer minimize integration efforts and manage all data flows.
associations, has built the MEXT Technology and Capability This approach has shortened the time to value for AI studies
Building Center to help its members and companies leverage due to the availability of clean data with real-time automated
digital transformation and technologies. The digital factory contextualization.
uses cutting-edge 5G technology on two production lines – a
discrete production line (end-to-end connected manufacturing) To accelerate the pace of the implementation of AI solutions,
and a continuous production line (digital twin of integrated steel an agile team was built comprising digital transformation and
production) – on which it has implemented over 160 use cases. business strategy professionals, data governance experts and
AI experts from member companies. Additionally, the in-house
MEXT is a real-world demonstration of the initiation phase of team has received training on artificial intelligence (AI)-specific
the guide. knowledge.

With a strategy to have a state-of-the-art digital model factory, MEXT has completed more than 150 smart industry readiness
leadership has prioritized AI-related initiatives and use case index (SIRI) assessments in the automotive, machinery and
development with a clear vision. The AI Lab for Manufacturing equipment producers, steel, textile, cement, chemical and food
is now being built and jointly works on various AI technology processing industries. Based on the SIRI insights, industry-
governance initiatives with the World Economic Forum’s Platform specific pain points have been defined with the structured
for Shaping the Future of Technology Governance: Artificial pathway of the AI Navigator. The AI use cases prioritized
Intelligence and Machine Learning and with public authorities. for implementation are workplace safety, AI-based machine
health monitoring and forecasting, quality inspection, process
MEXT is an openly available space to work with academics, optimization and predictive maintenance use cases. With in-house
manufacturing companies, technology providers and start- experts and upstream and downstream partners, additional AI
ups. It uses a Maestro smart data layer as a data governance use cases are being added and continuously updated.

Unlocking Value from Artificial Intelligence in Manufacturing 20

Conclusion
With a holistic approach, AI can solve some of the – Collectively pilot next-generation AI applications
most persistent problems in manufacturing and and unlock the untapped value of AI in
tap into new opportunities that allow companies to manufacturing, building on the capabilities of the
increase their operational performance, drive the technology centre, which has positioned itself as
sustainability agenda and empower the workforce. a testbed for industry and technology companies
that carry out research and development
While organizational and technological challenges innovations and proof of concept studies to
are still hindering the deployment of AI applications accelerate inclusive technology adoption;
at scale, leading manufacturers have successfully
seized the AI-derived potential and implemented – Foster a collaborative approach among a
a wide range of use cases for health and safety, diverse system of industry leaders, technology
quality, maintenance, production process, supply experts and academics to develop the right
chains, and resource and energy management. capabilities needed for AI deployment and
digital transformation in manufacturing.
By using a step-by-step approach such as the one
highlighted in this white paper, leaders can identify Manufacturing companies are invited to engage
relevant applications and successfully implement with the Centre for the Fourth Industrial Revolution
them. Türkiye and the Forum’s Platform for Shaping the
Future of Advanced Manufacturing and Value
Moving forward, the World Economic Forum and Chains and Platform for Shaping the Future of
the Centre for the Fourth Industrial Revolution Technology Governance: Artificial Intelligence and
Türkiye will continue to work closely with Machine Learning to collectively make progress on
stakeholders across industries and the Centre for the AI journey in manufacturing and further scale AI
the Fourth Industrial Revolution Network, offering capabilities globally, unlocking value for businesses,
MEXT as a unique testing and collaboration system workers, society and the environment.
for businesses to:

Unlocking Value from Artificial Intelligence in Manufacturing 21

Contributors
Lead authors

Ece Akın Armutak

Technology Governance Manager, Centre for the Fourth Industrial Revolution Türkiye

Memia Fendri
Initiatives and Community Lead, Advanced Manufacturing and Value Chains, World Economic Forum

World Economic Forum

Francisco Betti
Head, Shaping the Future of Advanced Manufacturing & Value Chains; Member of the Executive Committee

Felipe Bezamat
Head of Advanced Manufacturing Industry

Kay Firth-Butterfield
Head of Artificial Intelligence and Machine Learning; Member of the Executive Committee

Hubert Halopé
Platform Curator, Artificial Intelligence and Machine Learning

Vandana Menon
Engagement Lead, Centre for the Fourth Industrial Revolution Network and Partnerships

Centre for the Fourth Industrial Revolution Türkiye

Turkish Employers’ Association of Metal Industry

Tunç Acarkan
Director, Technology Governance

Efe Erdem
Executive Director, MEXT Technology Center and Centre for the Fourth Industrial Revolution Türkiye

Orkun Seçer
Additive Manufacturing Manager

Uğur Üresin
Fellow, Centre for the Fourth Industrial Revolution Türkiye

INC Invention Center

Benny Drescher
Technical Director, Artificial Intelligence

Toni Drescher
Chief Executive Officer

Patrick Kabasci
Director, Digital Operations Business Unit

Anne Loos
Executive Director, Artificial Intelligence

Unlocking Value from Artificial Intelligence in Manufacturing 22

Acknowledgements

The World Economic Forum and Centre for the Fourth Industrial Revolution Türkiye thank the following individuals
for their contributions through interviews, use cases and community discussions.

Çağlar Aksu Juergen Grotepass

Co-Founder and Chief Executive Officer, Sensemore, Chief Strategy Officer Manufacturing, Huawei, Germany
Türkiye
Uğur Gökdere
Serhat Murat Alagöz R&D and Software Development Consultant, Turkish
Data Analytics Manager, Tüpraş, Türkiye Employers’ Association of Metal Industry, Türkiye

Basmah AlBuhairan Pramod Gupta

Managing Director, Centre for the Fourth Industrial Vice-President, GEP Worldwide, USA
Revolution KSA, Saudi Arabia
Jay Lee
Meshal Almashari Professor and Founding Director, Industrial AI Center,
Head, Digital Design and Delivery, Saudi Aramco, University of Cincinnati, USA
Saudi Arabia
Humera Malik
Khalid Sulaiman Almudayfer Chief Executive Officer, Canvass AI, Canada
General Manager, Obeikan Digital Solutions,
Saudi Arabia Torbjørn Netland
Chair of Production and Operations Management, ETH
Özlem Altınışık Zurich
Intelligent Technologies Senior Director, Martur Fompak
International, Türkiye Jun Ni
Chief Manufacturing Officer, Contemporary Amperex
Şirin Altıok Technology (CATL), People’s Republic of China
AI-Data Science Chapter Lead, TOFAŞ, Türkiye
Yasemin Oral
Hande Bayrak Corporate Communications Director, Turkish
Project Management Director, Turkish Employers’ Employers’ Association of Metal Industry, Türkiye
Association of Metal Industry, Türkiye
Pamir Özbay
Patrick Brown Head, Operations, Fero Labs, United States
Programme Director, Predictronics, United States
Ersin Öztürk
Michelangelo Canzoneri Vice-President, Logistics, Digital Transformation,
Global Head, Group Smart Manufacturing, Merck Bosch, Türkiye
Group, Germany
Julian Senoner
Frans Cronje Chief Executive Officer, EthonAI, Switzerland
Chief Executive Officer, DataProphet, South Africa
Volkan Seyok
Güner Demirural Technical Manager, KARSU Tekstil, Türkiye
Smart Technologies and Innovation Lead, Ford Otosan,
Türkiye Anubhav Singh
Vice-President, Data and Analytics, Global Supply
Haldun Dingeç Chain, Schneider Electric, India
Executive Director, Digital Production Techniques,
Arçelik Global, Türkiye Murat Solmaz
Co-Founder and Chief Operations Officer, Maestro,
Çağkan Ekici Türkiye
Co-Chief Executive Officer, Khenda, USA
Daniel Szabo
Sercan Esen Chief Executive Officer, Koerber Digital, Germany
Co-Founder and Chief Executive Officer, Intenseye,
USA Yasir Tunçer
Co-Founder and Chief Executive Officer, Maestro,
Türkiye

Unlocking Value from Artificial Intelligence in Manufacturing 23

Metin Türkay Andre Yoon
Founder and Chairperson, SMARTOPT, Türkiye Chief Executive Officer, MakinaRocks, Republic of
Korea
Haydar Vural
Agile Tribe Lead, Data Science and AI Lead, Tofaş,
Türkiye

The World Economic Forum and Centre for the Fourth Industrial Revolution Türkiye would also like to thank the
following individuals for their contributions to the initiative and report.

Nilsu Atlan Jayant Narayan

Junior AI Project Specialist, Turkish Employers’ Project Lead, Artificial Intelligence and Machine
Association of Metal Industry, Türkiye Learning, World Economic Forum

Unlocking Value from Artificial Intelligence in Manufacturing 24

 Endnotes
1. McKinsey Global Institute, Notes From The AI Frontier: Insights From Hundreds of Use Cases, Discussion Paper,
April 2018, https://www.mckinsey.com/~/media/mckinsey/featured%20insights/artificial%20intelligence/notes%20
from%20the%20ai%20frontier%20applications%20and%20value%20of%20deep%20learning/notes-from-the-ai-frontier-
insights-from-hundreds-of-use-cases-discussion-paper.ashx
2. Loos, A., Sisejkovic, A., Drescher, B., “Humans at the Heart of AI”, The XXXIII ISPIM Innovation Conference “Innovating in
a Digital World”, held in Copenhagen, Denmark, June 2022.
3. MIT Sloan Management Review, Expanding AI’s Impact with Organizational Learning, October 2020,
https://sloanreview.mit.edu/projects/expanding-ais-impact-with-organizational-learning/
4. Lee, J., “Industrial AI: Applications with Sustainable Performance”, Springer and Shanghai Jiao Tong University Press,
2020.
5. McKinsey Global Institute, Notes From The AI Frontier: Insights From Hundreds of Use Cases, Discussion Paper, April
2018, https://www.mckinsey.com/~/media/mckinsey/featured%20insights/artificial%20intelligence/notes%20from%20
the%20ai%20frontier%20applications%20and%20value%20of%20deep%20learning/notes-from-the-ai-frontier-insights-
from-hundreds-of-use-cases-discussion-paper.ashx
6. BCG Platinion, Artificial Intelligence: Choosing the right approach to machine learning for your needs,
https://bcgplatinion.com/insights/ai-series-1-right-approach-to-machine-learning/
7. World Economic Forum, The Data-Driven Journey Towards Manufacturing Excellence, January 2022,
https://www.weforum.org/whitepapers/the-data-driven-journey-towards-manufacturing-excellence
8. Loos, A., Sisejkovic, A., Drescher, B., “Humans at the Heart of AI”, The XXXIII ISPIM Innovation Conference “Innovating in
a Digital World”, held in Copenhagen, Denmark, June 2022
9. Ibid.
10. Ibid.
11. Ibid.
12. Ibid.
13. Ibid.
14. Ibid.
15. Ludwig, H., Baracaldo, M., “Federated Learning: A Comprehensive Overview of Methods and Applications”, Springer,
2022
16. Loos, A., Sisejkovic, A., Drescher, B., “Humans at the Heart of AI”, The XXXIII ISPIM Innovation Conference “Innovating in
a Digital World”, held in Copenhagen, Denmark, June 2022

Unlocking Value from Artificial Intelligence in Manufacturing 25

 The World Economic Forum,
committed to improving
the state of the world, is the
International Organization for
Public-Private Cooperation.

The Forum engages the

foremost political, business
and other leaders of society
to shape global, regional
and industry agendas.

World Economic Forum

91–93 route de la Capite
CH-1223 Cologny/Geneva
Switzerland
Tel.: +41 (0) 22 869 1212
Fax: +41 (0) 22 786 2744
[email protected]
www.weforum.org
Unlocking Value from Artificial Intelligence in Manufacturing 26