WHITE PAPER

Democratizing AI:
Transforming Your Operating
Model to Support AI Adoption
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

AI for Everyone

When people talk about artificial intelligence (AI), the first

Adopting Artificial companies that come to mind are usually the FAANGs — ­
Facebook, Apple, Amazon, Netflix and Google. But this is far
Intelligence from a definitive list. Anyone can implement AI today, and the
FAANGs have no unique advantage.
Adopting artificial intelligence
(AI) has become a strategic The big technology companies achieved early successes with artificial
imperative. Early adopters are intelligence. Some even built their own specialized hardware, machine
achieving massive returns, learning frameworks, and research and development centers. These
while those who wait risk falling companies entered new markets, providing AI-led services and
behind. In order to empower products, often outcompeting the incumbents. They were rewarded with
more organizations and teams impressive growth.
to implement AI, it is essential to
embrace democratization, which AI has become a strategic imperative. Early adopters are achieving
puts the capabilities of AI into the massive returns, while others risk falling behind. Yet, with AI knowledge
hands of non-data scientists and and resources scarce, most companies cannot copy the FAANG
makes intelligence accessible to approach. Even if their vast resources were more readily available, it’s
every area of your organization. hard to reboot an organization without disrupting the bottom line. Most
Technological improvements help executives need to work with the resources available — their existing
achieve this vision, but it also
people, processes, and technology.
requires thinking about how an
organization is set up, from people Fortunately, any organization can implement and scale AI today. With
to processes. the emergence of enterprise AI platforms that automate and accelerate
the lifecycle of an AI project, businesses can build, deploy, and manage
AI applications to transform their products, services, and operations. As
the pioneer of automated machine learning, DataRobot has the wider
capabilities needed for an efficient end-to-end AI project, with a platform
supported by data analysts who integrated their business intelligence
and data visualization tools with AI in order to quickly transform
processes and decision-making.

However, democratizing AI within an organization requires more than just

software. It requires changing your operating model and finding the right
guidance to realize the full breadth of capabilities. This white paper provides
that guidance and takes a closer look at the steps you need to take.

1
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

Changing the Business Model

EVEN THE FAANGS DIDN’T START AS AI COMPANIES

Microsoft has restructured significantly over the last decade. Today, there is no Windows division. Everything
is increasingly centered around AI. Similarly, before 2015, Google used to make no mention of AI in its earnings
calls. Now it’s a core focus area. Netflix’s transformation is even more impressive. Having been founded in
late 1997 as a DVD sales company, it quickly pivoted into rental by mail. In 2000, it launched the streaming
service many know today and became an AI pioneer through its AI-powered, user-friendly interface and
recommendations. The now-defunct Blockbuster reached its peak in 2004, employing 84,000 people, but it
failed because it didn’t embrace the digital revolution. Its memory lives on in business school textbooks as a
cautionary tale about failing to adapt to technological change.

Why Democratize AI?

There are two reasons why democratization of AI is essential. First, there is a scarcity of experience in both data
science and AI implementation, so it is essential to bring in other resources to realize your organization’s goals.
Second, even if you can attract sufficient AI talent, adoption will only be successful with the support of the wider
business. Without the full adoption of AI practices, the scope and scale of your AI projects will be limited, if they
can be executed at all.

SCARCE AI TALENT
Demand for data scientists and data engineers has been rising exponentially. Research from the Royal
Society shows that demand has more than tripled over the last five years. LinkedIn’s analysis suggests
there’s a shortage of over 150,000 people with data science skills in the United States. Companies
pay large salaries in an attempt to lure data scientists to work for them. It’s no wonder that Glassdoor
repeatedly names “data scientist” as the best job in America.

2
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

The demand for AI talent is spread across all regions, yet skills are not so widely
dispersed. This makes the scarcity of talent even more acute for businesses with isolated
headquarters outside of major technology hubs and large cities.

What happens when you’re not in a location where there’s a deep talent pool for AI? Or, what
if you can’t afford the specialists? What can you do to avoid being left behind?

If you could put AI tools into the hands of more people within your organization,
imagine the impact it could have on your organization.

BUSINESS INVOLVEMENT
For any IT project to be successful, it should be aligned to business objectives. Organizations
that have proven most successful with AI take things one step further — getting their
executives on board to explicitly endorse AI as a key pillar of their strategy.

Once the tone is set from the top, the majority of the work of implementing AI is done within
the business itself. While the IT or data science team might lead the projects from a technical
perspective, it is essential that they involve the frontlines.

If the business doesn’t use the AI applications they’ve built, they have no practical value. To
avoid this fate and maximize impact, business leaders should help across the AI project
lifecycle. Business users can bring their domain expertise to frame the problem ­— setting
requirements, preparing data, reviewing results, integrating the AI application into their
processes, monitoring the AI’s outputs and supporting lasting business change. Domain
expertise is arguably one-third of the role of an ideal data scientist, beyond their programming
and mathematics skills. Yet, this skill is rarely comprehensive, particularly across any complex
business with a variety of products, services, and operational processes.

3
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

Required Skills
DOMAIN EXPERTISE
Domain Knowledge of the business
Expertise Knowledge of the data

PROGRAMMING SKILLS
Ability to write code to gather data
Ability to write code to explore/inspect data
Ability to write code to manipulate data
Data
Ability to write code to extract actionable intel
Scientist
Programming Ability to write code to build models
Math &
Skills Ability to write code to implement models
Stats
MATH & STATS
Foundational statistics
Internals of algorithms
Practical knowledge and experience
Knowing how to interpret and explain models

Data Scientists need a lot of skills and are scarce

The first step to scaling AI is identifying good use Change management should be an ongoing concern
cases. This requires framing the problem correctly to be to influence employee mindsets and encourage
solved with AI and then prioritizing based on potential positive behavior.
return on investment. Here, the business input is Transformation initiatives of any kind are challenging
indispensable, particularly if AI has been democratized and many fail. It’s essential to remember that
so that they can contribute in an informed manner. democratizing AI is not just about adopting technologies
that are more user-friendly, even if this helps. It is
Ideally, AI understanding is sufficiently ubiquitous
important to establish role models or “champions”
across teams so that they can independently identify
who demonstrate the benefits of using AI. People
new opportunities. To boost rapid opportunity
mimic those who surround them and can be inspired
identification, some organizations have turned towards
when their leaders behave differently. Staff ought
“crowdsourcing” mechanisms. One popular method is
to be supported in developing their skills so that
to run a periodic “AI challenge” across the business, with
they are confident in using AI. Leaders must foster
prizes for winners to encourage participation and pre-
understanding about why AI is being adopted and
funding from the executive to enable project delivery.
the value it can bring to organizations, in addition to
Another step to scaling AI is building trust. providing formal mechanisms to reinforce the adoption
A key part of building trust with business users is of AI, rather than hindering good behavior.
involving them in the implementation process. They
The starting points for most organizations looking to
are the ones who are best positioned to frame the
democratize AI are to invest in training and working to
business problem and define their requirements for AI
integrate AI with their legacy business intelligence (BI)
processes. They also typically have the deepest day-to-
tools. These are obvious first steps, but they are only
day understanding of the data being used, which means
part of the puzzle. In order to genuinely democratize AI
they can help accelerate an AI project and improve
and use it across a global enterprise, businesses need
results. It also helps minimize the risk that the AI models
to reevaluate their entire operating model, identifying
might be optimized for the wrong outcomes or using
opportunities to be AI-driven and removing barriers to AI
features that might be inappropriate from a business or
democratization.
ethical perspective.
There are six pillars to evaluate.

4
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

AI-Driven Operating Model

Vision & Strategy

Provide executive direction on how AI will drive
future organizational performance. Align AI to
existing strategic initiatives to maximize value.
Make difficult prioritization decisions to enable
focus. Encourage, empower and unblock change
across the business from processes to people.

Services & Customers ?

Determine what products and services you will
offer in future and to whom. This includes stopping
underperforming business lines, launching new
market leading AI-driven offerings, and and more effective
tailoring to customer needs via AI.

Processes & Channels

Automate work across the business wherever possible,
freeing up time for more complex challenges and richer
experiences. Use AI to enhance decision-making across
the remaining complex processes and customer channels. ? ?
Also establish a standardized repeatable process for
identifying and implementing new AI solutions at scale.

People & Organization

Set up an effective organizational model to scale
AI across the business while giving every employee
the opportunity to gain AI skills and experience.
Establish a Center of Excellence to provide
consistency, encourage rapid development, and share
best practices. Offer a new operating model that appeals to the
whole workforce, addressing job risk fears, while offering
more fulfillment and wider opportunities.

Governance & Reporting

Adopt and scale the use of tools that automate
and accelerate the lifecycle of AI projects to boost
data science productivity and enable democratic
contribution. Integrate your AI with existing investments
and complementary technologies, like cloud databases,
business intelligence tools, RPA (Robotic Process
Automation) and proprietary datasets.

Technology & Enablers

Create a governance framework that enables
customers, staff, and regulators to trust your AI,
such as ensuring that there are ethical guidelines
in place that fit your brand and values, removing
sources of bias, and implementing robust operational
controls. Use guardrails for AI models developed by
non-data scientists. Proactively monitor and manage
AI in production to maintain performance.

AI-DRIVEN BUSINESS
The AI-Driven Operating Model: Key Questions for Executives and

5
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

Questions Leaders Need to Answer

VISION AND STRATEGY

• Who will lead the AI initiatives?

• What targets will the executive team set?

• What changes to the composition or roles of the executive team are needed?

• How will leadership encourage, empower and unblock change?

• What “bargain” will we offer to staff that addresses job risk fears and offers the more
fulfillment and opportunity?

SERVICES AND CUSTOMERS

• What are the products and services we will offer? Which ones will we stop offering?

• What new AI-led offerings can we create?

• Who are our customers?

• Which customers should we target?

• How could we tailor our offerings more to individual customer needs?

PROCESSES AND CHANNELS

• Which existing process can be automated or enhanced with AI and through redesign?

• Which existing processes can be enhanced with AI and through redesign?

• What new processes will be needed to deliver our new AI-driven offerings and
customer centricity?

• How can we use AI to enhance our customer communication channels?

• What are our best AI use cases and which will be prioritized?

PEOPLE AND ORGANIZATION

• What capabilities will be needed across the organization?

• How should AI projects be staffed and organized?

• How many people are needed in each skill profile? (data scientists vs. business subject
matter experts)

• How do we roll out pragmatic AI education?

• How do we retrain and develop the wider skills of existing staff? What are our best AI
use cases and which will be prioritized?

6
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

TECHNOLOGY AND ENABLERS

• How will we prepare our data for AI?

• What technologies and platforms do we need to automate and accelerate

our AI projects?

• How will we integrate AI into our analytics and business intelligence?

• How will we integrate these technologies and tools with the wider tech stack?

• What other transformation enablers are needed? Do we need new equipment? Offices?

GOVERNANCE AND REPORTING

• How do we govern our AI programs and projects?

• How will we manage the operations and monitoring of our AI?

• How will we ensure that AI is trusted by our customers, staff and regulators?

• What ethical guidance will we provide?

• How will we document and report on compliance?

7
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

The Advent of Automated Machine Learning and Enterprise AI

Automated machine learning, which was invented by DataRobot, can create Enterprise AI
applications to address many of the challenges described above and make others more manageable.

Automatically creates
Automatically creates most
most of of Provides transparency into each Makes retraining models on
the documentation
the documentation required
required for model’s use of data, telling you new data and redeploying
for model validation and
model validation and model risk not just which features in the models into production simple,
model risk management
management — reducing the — data had the most impact to the fast, and low risk.
reducing the opportunity cost
opportunity cost of time spent predictive power of each model,
of
on time
lowerspent on lowerthat
value activities value
data but also explains individual
activities that data scientists,
scientists, almost universally, predictions down to specific data
almostspending
dislike universally,
timedislike
on. features and their values. Builds and ranks dozens of
spending time on. predictive models so you can
quickly evaluate and select
the model best suited to your
particular problem.
Reduces the cost, difficulty, and Provides tools for understanding
risk of deploying models into model accuracy and making
your production environment tradeoff decisions (e.g., between
Makes it easier to monitor model
by providing minimally invasive speed and accuracy, positive
performance and detect drift or
deployment options, such versus negative predictive value,
performance degradation over
as scoring code generation, when and where additional
time, alerting modelers to the
prediction APIs, and deployment models may be cost justifiable).
need for retraining or creation of
to Hadoop.
challenger models.

8
WHITE PAPER | Democratizing AI for All: Transforming Your Operating Model to Support AI Adoption

Conclusion
AI holds huge promise to help businesses transform. AI technologies are becoming
much easier to use, with substantial elements of data science automated or accelerated.
Democratization is essential to accelerate adoption and empower more organizations
and teams to realize its potential. This helps address the scarcity of AI talent and the
need for business involvement. However, democratizing AI successfully is not just about
technology. To genuinely democratize AI in your organization and reap the full benefits,
you need to consider embracing an AI-driven operating model.

9
 DataRobot is the leader in enterprise AI, delivering trusted AI technology and
enablement services to global enterprises competing in today’s Intelligence
Revolution. DataRobot’s enterprise AI platform democratizes data science with end-
to-end automation for building, deploying, and managing machine learning models.
This platform maximizes business value by delivering AI at scale and continuously
optimizing performance over time. The company’s proven combination of cutting
edge software and world-class AI implementation, training, and support services,
empowers any organization – regardless of size, industry, or resources – to drive
better business outcomes with AI.

Learn more at datarobot.com

© 2020 DataRobot, Inc. All rights reserved. DataRobot and the DataRobot logo are trademarks of DataRobot, Inc.
v04272020.0146
All other marks are trademarks or registered trademarks of their respective holders.