Applying AI:

How to find
and prioritize AI
use cases
“Don’t waste time on AI for AI’s
sake. Be motivated by what it
will do for you, not by how sci-fi
it sounds.”

Cassie Kozyrkov,
Chief Decision Scientist
at Google
 Applying AI

Introduction and
Motivation
Some companies truly struggle to define and prioritize
AI use cases. Good for them! The real problem arises for
the many enterprises that grossly underestimate the
challenge, feeling they have done the same thing many
times before. After all, everyone went through Digital
programs. Also, most have identified application areas
for Big Data. A common assumption is that the methods
for selecting successful AI use cases should not be all
that different.

The results are failed prototypes, but more often

great initial demos followed by a quagmire of disappo-
intments. For internal processes, AI deployment hits all
kinds of practical implementation hurdles. Alternatively,
scaling fails or the maintenance effort escalates. When
AI is applied in offerings, glitches turn customers off,
and the sales process draws out when customer data is
involved or there is no willingness to pay. In all cases, the
high hopes of value creation rarely materialize.

All of this might seem to be the inevitable and business aspects that were absent in
collateral damage of applying a disruptive both Digital and Big Data (without learning
technology. But it is not! A strong body of or even active decision making by a system).
experiences and deep expertise on the These critically influence the nature, the
subject is already available. With the proper value, and the ease of implementation of
preparation and understanding of the AI use cases. We support our analysis with
specific challenges of AI use cases and how instructive practical examples from in-depth
to address them, many of the above-men- experience on the subject in order to render
tioned developments can be anticipated and the subject as accessible as possible to the
avoided. business reader.

This paper shares critical lessons specific to Enjoy the ride!

identifying and prioritizing AI use cases and
provides a guide around the pitfalls. Fun-
damentally, the intrinsic interweaving of data
and learning algorithms introduces process

4
Applying AI

Elements of a comprehensive A company’s AI vision sets the high-level goals of any

AI strategy AI application to be developed or deployed. It includes
an understanding of the current position of the com-
pany, its competitive position and industry dynamics,
There is little doubt that AI will become relevant for all including potential changes to the industry’s business
companies, regardless of their industry or size. When model. On this basis, it can be decided where the orga-
it comes to creating value from AI, several pitfalls can nization could benefit most from AI − within a specific
be observed in practice – including the isolation of AI product or service and/or by improving processes.
use cases, the lack of resources and capabilities, and a The vision needs to be translated into a portfolio of AI
poor understanding of use cases and applications. use cases. To build this portfolio, you need to identify
To avoid this, a systematic approach towards AI is and prioritize relevant use cases.
needed. Therefore, from the very beginning, you need To execute the use cases, a set of enabling factors is
to be clear on the overarching objectives or purpose required concerning the organization, the people, the
of your company: What is its goal? Furthermore, it is technology, and the AI ecosystem.
necessary to understand how AI can help to achieve All of these aspects need to be taken into account
your objectives. when it comes to the development of a comprehensive
A comprehensive AI strategy consists of three parts: an AI strategy and are further detailed in our report “Ele-
AI vision, a portfolio of AI use cases, and a clear stra- ments of a comprehensive AI strategy.”
tegy for the required enabling factors.

AI strategy
aligned with overall strategy

Product/Service-centric AI Process-centric AI

AI
Vision
Competitive positioning

Ideation and prioritization

AI use
cases
Value chain configuration

Organization People Technology AI

ecosystem

Enabling Structure
Know-how
+ Talent
AI
Infrastructure
Startup
companies
factors
Culture
Governance Data Other partners
+ Collaboration

Execution

5
Applying AI

What is specific about

AI use cases?
We study AI use cases, which are defined as: a set of activities
taken to reach a specific goal from a business or customer
perspective, which involve a substantial application of artificial
intelligence. For the aficionados: we mostly consider the AI
subfield of machine learning — that is, applications of truly learning
algorithms. We draw no specific boundary on the sophistication
of the learning (i.e. “how intelligent” the application is). In our
experience, even simple learning might create high business value.

The overall process of defining and prioritizing AI use cases is de-

picted in the chart below:
Iterative approach

1. Ideation 2. Assessment 3. Prioritization Execution

Identify use cases Assess use cases by Cluster use cases Take use cases into
Preparation
aligned with AI vision value and complexity and prioritize production stage

Contents
• Understanding AI • Demand side methods • Structured assess- • (Iterative) Prioritiza- • Per use case: Define
• Maturity Assessment (Customer journey and ment of all identified tion of portfolio of scope for 1st data
• Business Priorities & process map) use cases use case ideas exploration and scope
AI Search Fields • Supply side methods • Assess dimensions • Based on ease of for MVP
(data and AI capabi- value and ease of implementation and • Data gathering
lities) implementation value • Model dev. & validation
Participants
• Business Top Mgt. • AI Strategists • AI Strategists • AI Strategists • Business Owner
• AI Strategists • AI Engineers • AI Engineers • AI Engineers • AI Engineers
• Business Owners/ • Business Owners/ • Business Owners/ • Business Owners/ • External Solution
Domain Experts Domain Experts Domain Experts Domain Experts Partners
• Data Experts/Owners • Data Experts/Owners • Data Experts/Owners • Data Experts/Owners

Output
• Good AI knowledge • Filled-out oppor- • Filled-out oppor- • Filled-out prioritiza- • Narrow-scope defi-
& relevant case tunity and solution tunity and solution tion matrix (value + nition for initial data
studies space templates per space templates per ease of implemen- exploration, broader
• Focus areas for AI identified use case identified use case tation) scope for MVP
• AI Maturity • Draft roadmap for development
prioritized use cases

6
 Applying AI

Let us summarize this and look at pertinent AI diction and optimization, but also creative
specifics. In the main body of this report we capabilities
will analyse and illustrate them in detail: • Finally goal-driven robotics/control (see
section 2)

Preparation For all these you need to understand: What

are table-stakes, what is hard, but doable,
Before you start, you should enable your what is expected to become available soon,
team and management with at least one and what has game-changing potential?
“Introduction to AI” workshop. This “demys- Data is the soil for AI algorithms and a source
tification and leveling of the playing field” is of inspiration in the ideation process — as well
crucial, as even in well-prepared companies as a source of frustration in assessment and
many business and functional leaders lack a implementation. Therefore, it is critical that
(common) understanding of AI. you conduct a review of sources for internal
and external data (structured and unstructu-
This introduction should be followed by the red), combined with options for data aug-
development of a top-down AI vision, sket- mentation, synthetic data, transfer learning,
ching out the areas where AI could add sub- simulated data, and more.
stantial business value — arguably, the most
common trap remains the a priori focus on Third, make sure you bring both parties to
search fields of marginal value. the table — deep domain expertise from the
business and in-depth AI expertise.
Finally, as a company, you need to assess your
AI maturity, including key enabling dimensi-
ons such as data and infrastructure, gover-
nance and culture, organization, and talent.
Step 2: Assessment
As we shall see, “AI maturity” is a critical input
for assessing to what extent more sophistica- Ultimately, we would like to prioritize AI use
ted applications can be implemented effi- case ideas based on their value and their
ciently and add value. A proper evaluation will ease of implementation. This requires asses-
help to preempt many of the most common sing these dimensions first:
failures.
• Ease of implementation: The difficulty
involved in implementing an AI use case

Step 1: Ideation depends on four categories: the required

data, the complexity of the required algo-
rithm, the required adoption of processes
When it comes to the ideation phase, three and systems, and the availability of the
points are important to spark great use case required know-how.Since the intelligence
ideas: of AI depends on data, often the most
difficult implementation issues involve
First, identify friction points or improvement data availability, quality, and updates.
opportunities in the strategic fields defined in Assessing this properly requires expe-
the AI vision — for example, along your custo- rience, knowledge in data management,
mer journey or process maps. We stress this and some sampling. For lower maturity
point since failing to target a business-critical levels, this absolutely requires the use of
problem remains, in our experience, a com- (independent) AI experts.
mon root cause for AI use cases not deliver- • Value: Assessing the value demands
ing value. domain knowledge/business owners.
In addition, there are a few AI specifics,
Second, complement your need identifi- which are rooted in the inductive learning
cation step by leveraging both AI capabili- and entanglement of data and algorithm.
ties and data sources. For the former, you’ll • For novel use cases or untested data
require an expertise of the true capabilities repertoires, there remains some uncer-
of AI. These capabilities include: tainty on the performance of the AI
application. While the expert knowledge
• Unstructured data, such as vision and lan- for professional estimates is quickly
guage, that can now be interpreted increasing, in practice we continue to see
• Novel analytic capabilities, such as pre- gross misjudgements. And even for the

7
Applying AI

top experts there remains the notorious Step 3: Prioritization

“unknown unknown” risk.
• More common is succumbing to the AI Prioritization seems trivial once a proper
paradox: It is deceptively easy to build assessment is done, and you may assume
successful AI prototypes, but fiendishly you can just plot the results and pick the
hard to scale AI — much harder, in fact, most attractive one. In practice, this is almost
than with traditional software or in Digital. never the case for AI use cases.
While one can view this as an “implemen-
tation” issue, it needs to be appreciated in First, the complexity of the assessment typi-
the value estimate: cally requires an iterative process where you
start with a rough assessment and prioriti-
• What scale (of possibly interacting AI zation, reassess what seemed attractive in
applications) is required to derive the more detail, conduct another review, and
envisioned value? then perhaps check red flags before you pro-
• What degree of maintenance is nee- vide a final go.
ded (such as adapting to new chan-
ging data-sets/environments)? Throughout this process, you’ll almost always
discover two things:
Forgetting these facets is a common
cause for naively overestimating the • All your high-value use cases might have
achievable value. high complexity — so you try to decom-
pose them into intermediate “viable
One aspect combining value, ease of imple- products” and approach the roadmap in
mentation, and a few strategic considerati- an agile way.
ons is AI make-or-buy decisions. There are • Even more common is that your use
some distinct characteristics and criteria cases, while well-separated on the value
when suppliers require company data to train axis (by definition), have technological
an algorithmic solution — from IP/risk issues interdependencies. Clustering those can
to contract management and collaboration, drastically speed up the collective imple-
all the way to maintenance. We discuss this in mentation, while concurrently de-risking
some detail in an insert below. the approach.

Finally, some AI flags should be set at this

stage: They concern important implemen-
tation aspects that can increase the effort
Execution
(or decrease the value) substantially and
might require a separate “quick assessment” While execution is not the focus of this report,
request. we will comment on what aspects of the
• Regulatory and/or ethics flag - mostly assessment and prioritization process require
when sensitive people-related data and proper documentation and how they might
or decisions are involved influence the set-up of the execution.
• Risk and transparency issues. These
include: We will cover all the above points in the ensu-
ing sections, but the summary may already
• Cyber-security flag: Truly critical allow for an initial understanding of the most
when processes are fully automated important specifics of AI and sources of com-
and might get infiltrated mon failures quoted in the introduction.
• Black swan resilience/human-in-the-
loop flag: When extraordinary events
could lead to severe malfunctioning
(e.g. financial trading algorithms
facing COVID-19)

8
 Applying AI

Preparation
AI is both the most significant business transformation
force and the largest business opportunity driver we
have seen in a long time. As a consequence, mana-
gers should be aware that a programmatic approach
to leverage AI for their company encompasses many
dimensions, which is summarized in The AI Strategy
House.

Use case ideation and prioritization is a key Second, there is a bit of a chicken-and-egg
element of such a programmatic approach. challenge in requiring a top-down AI Vision
In order to succeed, thorough preparation is for defining the broad areas where AI could
required. add significant business value, and the need
for some use case experience to actually
First, you need to ensure everyone has a do so properly. Since a common trap conti-
good understanding of AI — its power, but nues to be teams wasting time on marginal
also its unusual properties — and that the AI application areas, senior business lea-
key areas of potentially high business impact ders — building on an initial understanding
have been defined. An Introduction to AI of AI — should make the effort to define and
workshop is mandatory, to ensure a common document the key areas where they expect
understanding and alignment of business, value and why. It may be in specific products
functional, and technical contributors on the and services, some core processes (such as
team and beyond. supply chain or customer-facing activities), or
support processes (such as finance and HR).
This introduction should review basics,
demystify the subject, provide concrete This vision serves as a business guideline for
examples from your industry and beyond, the use case ideation process, even if it might
and describe typical hurdles and pitfalls in be revisited based on experiences gained
a business context. This workshop can be during the process.
held at different levels of sophistication, and
for more mature companies becomes more Specifically, there is one scenario where such
of a joint review of AI, including examples of an AI Vision top-level view is decisive: when
already successful company implementati- there is the potential of a true business dis-
ons and experiences. ruption via AI. Industries, such as finance and
insurance, healthcare, automotive/mobility
When we at appliedAI were approached by or media — to name a few — clearly have such
a midsized company to help them identify exposed domains, which aggressive disrup-
relevant AI use cases within their automation tors are trying to exploit. Addressing such a
business, we first had to convince them to fundamental business model challenge via AI
start with such a training as they expected requires a separate approach, as it is rarely
us to bring in the necessary AI expertise. derived bottom-up from individual use cases.
This training proved to be the decisive suc-
cess factor for the use case workshops that Finally, companies are at drastically diverse
followed, as it ensured that everyone involved levels of AI maturity. We at appliedAI have
— from the managing director to the domain developed a comprehensive AI Maturity
experts — had the same understanding of AI Assessment, based on an efficient online sur-
so that concrete use cases could be determi- vey, that allows companies to quickly assess
ned and discussed quickly. themselves regarding AI. It covers all the ele-

9
Applying AI

ments of the AI Strategy House (see sidebar). zation and talent, governance, risk mitiga-
tion, and ecosystem management. Lacking a
We strongly recommend that companies comprehensive view of maturity will lead to
complete such an assessment before embar- highly inefficient and failure-prone processes
king on prioritizing use cases, as the feasibi- of estimating the ease of implementation.
lity of more sophisticated applications criti-
cally depends on the maturity in the enabling With this, you are ready to embark on use
factors — AI infrastructure and data, organi- case ideation.

Ideation: Identifying
relevant cases
The ideation phase aims to identify new and relevant AI
use cases within the organization. It requires bringing
together deep domain expertise from the business and
in-depth AI expertise. When the latter is not available
from internal AI Centers of Excellence, leverage exter-
nal expertise — but make sure that you use the process
to strengthen the internal competence. To make the
best use of scarce team time, use case ideation should
follow an efficient process.

Your company’s AI vision (see previous sec- med process is an insurance company
tion) defines what areas to focus on to create using image recognition to automatically
maximum business value. At a high level, we assess car accident claims with limi-
distinguish two directions: ted human supervision. This drastically
lowers the cost of handling claims, while
also saving time, thus boosting customer
• For process-centric AI use cases, the experience.
focus lies on company processes, either Support processes in finance, HR, IT,
internal or at the interface with custo- and more should not be neglected. Truly
mers, partners, or suppliers. Often, the strong organizations excel in them. Often,
competitive value lies in company core these processes allow the partial lever-
processes. Start with those that are age of external suppliers (see make-or-
costly and/or important to your business, buy below) and might thus trigger some
then ask: Can this be automated? Could quick wins.
it be faster or more precise? Keep in mind • Product/service-centric AI use cases are
that AI can be used either to support often more challenging in their definition,
existing processes or create new ones. as they require deeper technical exper-
One example of a disruptively transfor- tise and customer experience. They may

10
 Applying AI

profit from including design/UX experts for the This approach helps you develop solutions that actually
ideation session. On the other hand, compared to solve existing user problems or daily challenges and,
process-centric cases, AI in company offerings thus, increase their adoption rate. Ideally, the solution is
demands less change management later in the intuitive and easily integrated into the customer’s daily
implementation. habits.
The focus is of course on those areas with the
highest revenue/profit potential and/or a strategic With the spread of AI, still unconventional situati-
importance. Needs and friction points in customer ons may arise where the users of your offerings are
journeys are at the core: How can AI be used to machine-based processes and decision makers, not
improve existing products or create new ones that humans (e.g. in production environments or finance).
address unmet needs of your current or potential Getting access to the friction points in such processes
customers? Entirely new AI-driven offerings can be is typically much more challenging, requiring access to
generated by assessing potential uses of AI. company systems.

The input for identifying use cases is somewhat similar Process Map
to traditional ideation sessions on the demand side (for Use a process map to systematically break down
example, through process maps and customer jour- your processes into tasks and decisions, then identify
ney/friction points), but is highly AI-specific on the sup- individual elements that can be automated using AI.
ply side (AI capabilities and data maps). Let us discuss Remember, you want to consider automating tasks, not
them sequentially. jobs.

AI can be particularly beneficial when there are mul-

Demand Side Methods: tiple changing variables that determine a decision. As a

learning system, AI can automate decision support and
Customer Journey Map actions with complex dependencies on dynamic data,
The customer journey map is helpful for visualizing which can no longer be captured in traditional “rules.”
each step of the customer interaction and use of the For that, ideate in fields where AI can improve the ove-
product or service. The method involves considering rall process by eliminating pain points — or where you
the customer experience from the point of view of can redesign an entire process with the use of AI.
a specific personas (to force you to be very specific
about your “customer”).

For this approach, map out each persona’s main points

Supply Side Methods:
of interaction, and identify critical sections that could AI Capabilities
be improved with AI solutions. Additionally, actual Correctly applied, the most important novel input for AI
observations of the target group and personal inter- use cases comes from studying the business-relevant
views with users or external experts always deliver capabilities AI has developed. As an immediate caveat,
novel insights with respect to customer pain points. AI capabilities that are naively applied, following hype

Computer Vision Computer Audition Computer Linguistics Robotics and Control

Process visual data and recog- Process and interpret audio Process, interpret, and render Analyze, interpret, and learn
nize objects Understand the signals text and speech from data representing physical
semantics of images or video systems (incl. IoT) and control its
sequences behavior

AI Capabilities

Make predictions about future Process large amounts of data Look for optimal solutions to Generate images, music,
course of time series or likeli- and find patterns and “logical” problems with large solution speech, and more based on
hood of events relationships space sample creations

Forecasting Discovery Planning and Search Creation

11
Applying AI

reports or vendor marketing, have also been learning and simulating their environment,
the source of sad flops. Thus, we at appliedAI trading off exploration and exploitation.
have put significant effort into structuring
these areas and making them accessible to For all these capabilities it is truly critical to
business people. understand precisely: What are table stakes
today? What is doable, but requires sub-
Overall, we distinguish the following eight stantial effort (and data)? What is expected
capabilities. The first three are capabilities to become accessible over the next couple
that allow unstructured data to be used: years (e.g. in vision/speech)? What could be
• Computer Vision game changers (in particular around crea-
• Computer Audition (including speech to tivity and goal-based control/simulation)?
text) What is simply irrelevant science fiction?
• Computer Linguistics (chatbots, trans-
lators, etc.) Understanding these points will also be cri-
tical in make-or-buy decisions and vendor
These are critical ancillary capabilities that selection (see insert). We at appliedAI conti-
allow machines to act in the real world and nuously educate our partners in these areas,
interact with humans, and are often used as and have developed AI application playing
advanced input for further processing. cards for illustrative purposes and as a great
tool for ideation workshop settings.
Second, there are new analytics capabilities:
Data Map
• Forecasting The other critical supply-based input for AI
• Planning/Optimization use cases is data.
• Discovery (pattern recognition)
First, you should understand all possible
These are at the core of many business appli- sources of data — internal, external, struc-
cations. Wherever you still use rules-based, tured, and unstructured. You should also
linear, or manual approaches for forecasting be aware of methods to generate data
and optimization, AI methods are a candy (augmentation and synthetic data, but also
store for quantum leaps in performance. Also, simulation and generative methods). Finally,
look for areas where learning can be cen- there should be some understanding of the
tralized while keeping actions decentralized option to learn from related data (aka “trans-
(from next-best action in sales to naviga- fer learning”).
ting trains/planes/ships/cars). Lastly, check
for areas where speed is of critical value Building on this, you would start by asses-
(machine handling, financial markets, etc.), sing your company’s potential data assets
as AI systems have the potential to be several and options for complementing them. Then
orders of magnitude faster than humans. determine which of the data assets are uni-
que and could be the source for potential use
Finally there are truly advanced capabilities: cases. Finally, elaborate how you can apply
the data in a useful way internally or for adja-
• Creation cent businesses.
• Advanced Robotics and Control
You may also already want to consider how
There is a lingering belief that machines are accessible the data is today and how that can
not capable of creating new content. Howe- change over time. This will be a major focus in
ver, the last years have seen the rise of new the assessment section.
generative methods, involving “adversarial”
algorithms - meaning they generate new Obviously, all such demand- and supply-side
input data from target outputs, using two input areas should be complemented by best
interacting algorithms. Such applications practice use case examples from within your
constitute an -albeit still embryonic - encro- industry and beyond. We at appliedAI provide
achment upon traditional domains of human a comprehensive use case library and con-
creativity. tinuously collaborate with our industry and
national/international institutional partners to
Another truly sophisticated area is Advanced grow this repository.
Robotics and Control, applicable primarily in Combining all the above in a concrete setting
industrial settings. It often involves agents builds on classic brainstorming techniques

12
 Applying AI

that are often used in a workshop setting. For this, you The outcome of the ideation phase should be cap-
should prepare content in advance for your top pro- tured in opportunity and solution space cards. Utilize
cesses and products — data maps, costs drivers, reve- the opportunity space card to identify the issue to be
nue share, etc. — and share that content with the group. solved with AI (which product, service, or process is
Carefully select the right mix of participants, including affected by the use case and the user story), and ask
people familiar with AI technologies, data infrastruc- the following questions: What is the objective of the
ture, and existing AI initiatives, as well as people familiar use case? Where does AI contribute? Which products
with the company’s products and services. or processes could it be applied to?
Once again, the level of detail is important. “Introdu-
As in all ideation sessions, make sure to follow the idea- cing a chatbot” is not a helpful description of a use
tion rules in the workshop and defer early judgement case. The following would be more useful: “Use AI to
— you will have plenty of opportunity for that later. We reduce customer service costs by predicting custo-
at appliedAI have a wide set of offerings to successfully mer intentions and automatically handling the most
facilitate such ideation workshops. frequent requests via a chatbot interface in the online
customer service process.”
When a European energy supplier set out to find AI use
cases, they were faced with the challenge of the “AI After that, use the solution space card to indicate the
divide”: On one the hand, there was a central AI team of capability used to solve the issue, the desired output of
experts who understood AI‘s capabilities. At the same the solution, and what information or data is needed to
time, however, the problems were „trapped“ in the train AI. Remember that a solution should always solve
various business areas by the domain experts. To over- a specific problem.
come this gap, a series of use case workshops was set
up with all the different business units, including time
allocated for providing an introduction to AI. Although
this is a long and costly process that requires the coor-
dination and support of a large number of participants,
it is necessary for good results.

13
Applying AI

Assessment:
Assessing use cases
Once you have a good list of potential use cases, you
are ready to address the second challenge: How to as-
sess what a truly valuable and feasible use case is. To do
so, you will need to evaluate each of the ideas on your
list, starting with a precise description of each case. You
need to decide how critical the use case is. Does it ent-
ail a high level of risk? Does it require extreme accuracy
(e.g. for autonomous driving)?

At the core, you need to make an assessment AI application. This means you need to
of both the value and the ease (or comple- assess the value depending on what
xity) of implementation of a use case. Addi- accuracy can be achieved — or, alter-
tionally, check a few common flags. For this, natively, define minimum accuracy
you can use the appliedAI AI Use Case Can- levels required.
vas. The following aspects should be asses- You should also be aware of the AI paradox:
sed: It is deceptively easy to build successful AI
prototypes, but fiendishly hard to scale AI —
much harder than with traditional software
Value or Digital. This applies to the build phase, but
even more to maintaining potentially inter-
• Business value: What is the economic acting learning applications at scale in an
potential of the use case in terms of cost environment of dynamically changing data.
reduction, additional sales, speed-up, Thus, it is critical to define:
quality improvement, or customer satis-
faction? • What scale (of possibly interac-
• Strategic alignment: How does the use ting AI applications) is required to
case fit into the strategic goals of the derive value?
company, and what advantage does it • Over what time period is the value
bring in terms of strategic market positio- derived?
ning, and improved resource efficiency, • What dynamic changes does the
etc.? application need to adapt to?
In principle, these dimensions are metho- • What degree of maintenance is
dologically straightforward and familiar, needed?
from business decisions in general. There
are, however, some AI-specific challen- Forgetting to answer these questions
ges that mainly result from the entwine- are standard value-destruction traps.
ment of data and algorithm in the induc-
tive learning process:
Ease of implementation
• For novel use cases or untested data
repertoires, there remains some Estimating the effort required for each AI
uncertainty on the performance of the use case is notoriously difficult without an

14
Applying AI

15
Applying AI

in-depth knowledge of the quality and avai- internal data and the resulting algorithm
lability of all required data over time and full contains critical business knowledge, deci-
information about the AI maturity. The ding on, contracting and managing the relati-
appliedAI Use Case Canvas offers a frame- onship has some complex, unique issues. We
work to derive a solid estimate. discuss the topic in more detail in the sidebar.

You need to evaluate four main categories Finally, there are some AI flags that should be
(data, algorithm, process/systems, and set at this stage. These are vital implementa-
know-how), while also estimating the dura- tion aspects that can turn out to increase the
tion, until you have a fully working and ver- effort (or decrease the value) substantially
ified. Aggregating those provides a score — and might require a separate “quick assess-
for the overall ease of implementation. The ment” request.
higher the score, the easier it should be to
develop. • Regulatory and/or ethics: This is mostly
relevant when sensitive people-related
An AI expert with sufficient knowledge of the data and/or decisions are involved. For
product or process related to the use case is instance, a regulatory authority might
typically required for the assessment. Gene- require transparency of the algorith-
rally speaking, the more high-quality data mic decision making — often in terms
you have, the less explicit process know-how of outdated metrics stemming from
you will require in the implementation phase; rule-based decision systems. Or there
in the assessment phase, such know-how is might be some legal or consumer-driven
indispensable. “fairness” expectations (often with only
vaguely defined criteria). Since AI-based
Below is a short description of the respective decisions can happen at unprecedented
categories: speed and scale, while at the same time
everything is measured, such topics can
• Data: What data do you need for the use become of enormous importance and
case? What quality of data do you have? need to be flagged in a value assessment.
What effort is required to integrate the • Risk and transparency: There is a wide
relevant data and make it available? What range of further risks that need to at least
dynamic updates are required over time, be documented in the assessment.
and how can they be performed?
• Algorithm: Are there other implemen- • Cybersecurity is becoming increa-
tations of the use case already up and singly important the more we digita-
running, either within your company or lize our businesses. Most executives
elsewhere (including other companies or think of data security, but automating
industries)? with AI introduces a new quality of
• Processes/Systems: Which processes risk: process security (hackers have
and systems are affected? Do you need famously demonstrated the take-over
to make major changes to existing pro- of an autonomous car to make such
cesses or systems? a risk tangible). High-risk areas thus
• Required Expertise: Do you have the require an extra cybersecurity flag.
necessary technical and domain know- • Resilience: Depending on the envi-
ledge? sioned lifetime of the use case, it
may be important to expose it to an
The AI Maturity Assessment performed in the extensive test of resilience against
preparation phase will help you tremendously future extreme events not represen-
in this evaluation. Often, a rough first assess- ted in the test sets. Generalization to
ment is completed in a workshop setting and such events has not been assessed.
validated in detail afterwards. But be aware This requires algorithms that at least
that there remains an “unknown unknown” recognize they are on uncertain ter-
risk that might only surface in the PoC phase. ritory, often combined with human-
in-the-loop systems. Again, a quick
One aspect that affects value, ease of imple- check of such an exposure should be
mentation, and a few strategic considera- performed and potentially flagged.
tions is AI make-or-buy decisions. Since AI
requires training, it is not plug-and-play. Spe- With a robust assessment in place, we are
cifically, when suppliers require access to finally ready for prioritization.

16
 Applying AI

Prioritization:
Clustering and
prioritizing use cases
Once a set of use cases has been identified and asses-
sed, you still have to make sure you focus on those that
are most relevant — after all, the potential applications
of AI are endless. In order to apply your AI competence
most effectively, the use cases must be prioritized.

The basic approach involves plotting the value of use cases is separable (this is the
results of your assessment phase in a prioriti- way we defined them) there may be cluster
zation matrix and choosing the most attrac- synergies that strongly affect the collective
tive cases. Here, there are several considera- ease of implementation.
tions to take into account.
Thus, it is useful to cluster use cases by the
First, the process is not as linear as it seems. following parameters:
Assessing use cases is quite complex, as we
have seen. Typically, there is a first round of • The required input data
rough assessment and prioritization. Then, • The AI capability it will require
a deep dive on the most promising looking • The product(s) or process(es) to which it
cases, followed by a second round of review could be applied
and prioritization.
Prioritize the clusters that have a large num-
In addition to this iterative approach, you will ber of use cases with high value, and at least
typically face two additional issues: Often, some use cases that are relatively easy to
many high value use cases seem to face implement within a previously determined
substantial implementation hurdles. It is specific timeframe. Within the prioritized
worthwhile to assess to what extent these clusters, pick one or two cases for validation.
challenges can be resolved by breaking down Ultimately, this will mean you are dealing with
the use case into smaller, more digestible a total of about three to six cases in one wave.
units — each with well-defined value steps —
and approaching the resulting roadmap in an Good candidates can have different charac-
agile way. teristics. For example, it might be possible to
implement some of them quickly and realize
A second issue occurs when use cases are “quick wins.” Others might have a high stra-
technically interdependent. AI capabilities tegic value or marketing relevance. Or they
such as computer vision or natural language could be easily scalable to other products
processing might be shared across use or processes. Ideally, these clusters are the
cases, and the same can hold true for data base for building your use case roadmap.
assets and pipelines. These often require
substantial investments in skills and infras- So far, we have rarely seen companies that
tructure to resolve. In other words, while the consider the dependencies of their AI use

17
Applying AI

cases. A notable exception is a major European manu- Besides to centrally driven prioritization and roadmap
facturing company that systematically considers the building efforts, there will usually also be some instan-
reusability of an AI model for future use cases as a key ces where decentralized units will drive their own AI
factor for prioritizing use cases. solutions with their own resources and budgets. This is
most often the case when the use case may not end up
Prioritizing use cases works differently in compa- high on a company-wide prioritization list but will solve
nies that are just getting started with AI (i.e. “experi- an urgent, unit-specific problem.
menters” in the maturity assessment). Here, it is espe-
cially important to manage the risk of a first use case, Clear guidelines are required for such decentral
although this may not yet be the most valuable one. A activities. Typically, priority depends on the longevity
good choice might be one for which data is available of a use case and the likelihood for it to interact with
and a similar solution already exists in another part of other company processes in the future. For instance,
the organization. distinct efforts for forecasting sales in business units
will lead to escalating complexity when a consolida-
When mobilization is an important goal, use cases tion of numbers is required. Also, at the very least, data
should also be “visible”. Most likely the use case needs governance and platform decisions should be coordi-
to be related to the core business. Visibility can be nated holistically — such coherence is critical for being
enhanced by the internal communications depart- prepared for applications accessible at higher maturity
ment, which should also ensure that the technical levels.
language is comprehensible and the value it adds to
the business is clear. Is the goal to address a specific Decentral entrepreneurial AI activities should be
process? Or to explore new fields of technology? Set encouraged (or at least tolerated), but coordination of
expectations accordingly. the above-mentioned dimensions needs to be assured
and the company’s overall AI vision should be respec-
ted at all times.

Iterate as necessary

Assessment of Use Case Ideas Prioritization of Use Case Ideas

Dimension Specification
High

Economic value
Value
Strategic alignment
Value

Data
Ease of
implemen- Algorithm
tation
Process/Systems
Low

Required know-how Low Ease of implementation High

Prioritization of identified ideas along core • Assess use case ideas along core dimensions of
dimensions of value and implementation value and implementation effort (incl. AI specifics)
effort (incl. AI specifics) • Decompose very strategic, complex Use Case into
smaller solutions to make them manageable as
intermediate viable products
• Cluster use cases with similar underlying techno-
logy (capabilities/data)

18
 Applying AI

Make-or-Buy Decisions
for AI Use Cases
Most organizations that want to apply AI do not think
of themselves as technology companies and would in-
tuitively try to find a supplier to “buy” from. Some even
recall painful experiences from the early days of en-
terprise software applications. At first, all applications
were approached as customized projects only to run
into extensive maintenance issues. Until, finally, the relief
of productized enterprise software arrived. The cur-
rent state of AI implementations would seem to exhibit
some parallels.

That may be true, but there are critical agreement on data access, confidentiality,
differences — most importantly, that “make” and more — and in a business context,
and “buy” acquire quite a different meaning the required adaptation is typically quite
in the context of AI (i.e. inductive learning extensive.
systems). The vast majority of raw AI
algorithms are (still) open source, available An area where many companies have started
for free, but without immediate business to experience the fluidity of the make-or-
value. Only after training on data (often at buy “decision” is chat bots. Most companies
least partially your own company data) does cannot — and should not — build them from
the trained algorithm exhibit intelligence. scratch. On the other hand, if you ever tried
This is quite distinct from the mere to just implement a pre-trained version
parametrization of traditional enterprise without further learning, your experience
software. An executive once compared this must have been terrible. The best vendors
to new hires requiring training to provide offer pre-trained models with graphical user
business value. You can shorten the time by interfaces to support the further process of
hiring them “pre-trained” by schools and customization (i.e. continuing to learn). This is
universities, but they will still need to learn a general trend we are seeing in the industry
your specific company environment. — for simple learning processes, the AI “build”
process is being supported by ever more
Similarly, machine intelligence is not plug- accessible tools, lowering the requirements
and-play. In AI, make-or-buy is more of a for technical expertise.
continuum than a binary decision. At one
extreme, even when taking full ownership, The correct reformulation of make-or-buy is:
your teams will never build algorithms from To what degree do you want to do it your-
scratch. They will always use packages, often self vs partnering in regards to building and
pre-trained on some AI capabilities. At the managing the AI application? And with whom
other extreme, even the most productized should you partner? And how should you
AI (think of a spam filter) needs to adapt structure the partnership or contract?
to you, so there needs to be at least some

19
 Applying AI
High

Protect your core business Immediate sweet AI spots

Strategic value potential

Build distinctive capabilities Do AI yourself and scale fast

vs competitors

(e.g. acquihire)

Outsource as much as possible Potential AI business opportunity

Focus on contract management Explore, whether high value to other
parties
Low

Low Unfair Advantage in AI Build (data, skills) High

Criteria:

First, what is the strategic value for you? (Remember,

you can only excel in what you own.)
Second, Can you build it? (Your AI capabilities and your
preferred access to data).

A general consensus is that you should refrain from

building your own AI solutions where ERP providers
(SAP, Salesforce, and Oracle) or process specialists
(such as ServiceNow or Workday) are already integra-
ting AI capabilities of the data handled in their systems
into their suite. You are towards the lower left side of
the matrix (at a disadvantage for building the AI solu-
tions) and, often, these are not at the core of your com-
petitive differentiation.

Further details of make-or-buy deserve a much dee-

per discussion and are the subject of a separate publi-
cation where we also address:

• How to choose among major vendors, start-ups,

academia, consultancies, etc.
• How to structure contracts: IP, confidentiality,
maintenance, regulatory, and performance gua-
rantees

20
 Applying AI

Execution:
Getting started
Before you move to the implementation of a prioritized
use case, there is one last thing you need to do: dou-
ble-check your assumptions, both from a business and
technical perspective. Make sure you fully understand
the characteristics driving the value of the use case (le-
vel of accuracy, scale, and maintenance requirements).

Do you really have the data, and is it accessible via so-

lid pipelines? What red flags does your use case have
(regulatory, ethical, risk and transparency), and are you
sure you can address them?

Finally, are you truly clear on the make-vs-buy decision:

Do you want to develop a case internally, find a partner
to develop it, or buy a solution? It is also good practice
to define gates and knock-out criteria that would make
you abandon the pursuit.

Now you are ready to start execution. Implementing

AI use cases has its own body of experiences of best
practices and traps, which we will cover in a future pu-
blication.

21
Applying AI

Our partner Siemens

is finding use cases in
5 days
Conducting a use case ideation in an established organization is
often challenging, as the necessary employees are typically lo-
cked into their routines and other projects. Our partner Siemens
has chosen a unique approach to address these problems: The
Siemens AI lab and its 5-day use case sprints. In the following they
share their experience.

As a leading provider of industrial solutions, Siemens open innovation platform where business stakeholders
is researching and developing AI technologies. “Our and their domain experts are matched with internal AI
goal is to create value for our customers through AI, to researchers to find suitable solutions together.
improve our internal process efficiency, and to make a
positive impact on society and our environment.” With an entrepreneurial mindset and a fail-fast culture,
teams develop tangible prototypes for the next game-
Founded in 2017, the Siemens AI Lab represents a changer in a format spanning less than a week. The Lab
platform that drives the progress and adoption of AI provides tools and methods to bolster the potential of
technology with meaningful impact for Siemens. AI across multiple business units, with the aim of disco-
vering scalable applications for the company.
The Munich-based Lab explores the potential of emer-
ging AI technologies for industrial use cases. As part of Siemens has adopted the format elsewhere, including
the Data Analytics & AI research hub of the Corporate Beijing and Berkeley.
Technology division, the Siemens AI Lab acts as an

The offering

Picture:
22 The Siemens AI Lab engine with its key elements
 Applying AI

Acceleration The Siemens AI Lab brings the right people together in

Under the motto “fail fast in five days” the Siemens AI the right place and complements the customer’s data
Lab teams up with internal customers and AI experts and domain expertise with industrial experts as well as
to build a proof of concept, a prototype, or a minimal proven AI technologies.
viable product quickly and affordably.
Lessons learned:
The Lab supports the project team throughout the 1. Be there from the start.
whole process. For example, it grants access to over Support the shaping of project goals and setup from
100 AI experts from Corporate Technology, which the very beginning.
offers the perfect starting point for a detailed assess- 2. Co-locate the team.
ment of any use case hypothesis and the required No distractions, all the expertise in one (physical)
dataset. In order to prepare project teams for the room.
intensive Acceleration sprints, questions, details, and 3. Fast go/no-go decision.
the expected results are clarified beforehand. Implement and test a Proof of Concept idea in 5
days.

Orientation

Picture:
Value Proposition Canvas adopted to AI specifics
23
Applying AI

At the Siemens AI Lab, customers can learn in 1. Holacratic team organization

joint teams about the most impactful Indust- 2. Open-mindedness
rial AI use cases and AI application to improve 3. Agile project management
their business.
Instead of having a job description to adhere
Ingredients for a successful workshop to, the team members at the Siemens AI Lab
1. Start with the introduction of Industrial AI are organized by purpose-driven roles that
Introduction & Tech Session correspond to the “why” and “how” instead
2. Have a look at the Customer Situation & of the “what”. Being open-minded is not just
Data Value important for the sake of the work, but also
3. Get inspiration from Abstracted Use Cases for how colleagues collaborate.
4. Use Case Ideation & Design Thinking
5. Adapt it to the Customer Situation & Data Our goal is to expand the Siemens comfort
Value zone in AI by building a sustainable portfo-
6. Create the Action Plan & Way Forward lio of skills and tools that drives AI@Siemens
towards a new innovation culture, where
Exploration diversity, agility and openness are the buil-
Piloted in early 2020, the Residency Program ding blocks of corporate intrapreneurship.
offers promising talents in applied AI the
opportunity to drive Industrial AI research. In
this 9-month program, participants solve fun-
damental challenges of Industrial AI and have
access to a broad network of researchers
within Siemens.

“Culture eats strategy

for breakfast.”

Whether it is acceleration sprints, orienta-

tion workshops, or exploration research, the
Siemens AI Lab only succeeds thanks to its
innovative culture. Three factors are key to
ensure the success of AI Lab:

24
 Applying AI

25 25
Authors Contributors
Hendrik Brakemeier is Senior AI Strategist at appliedAI. The authors would like to thank the SiemensAI Lab,
Before joining appliedAI Hendrik pursued a doctoral namely Vera Eger and Bernd Blumoser for their contri-
degree at the software and digital business group at TU bution to this paper.
Darmstadt during which he worked on the economics
of data-based business models. Furthermore, he Bernd „Benno“ Blumoser is the Innovation Head of
served as part of Accenture‘s analytics strategy and the Siemens AI Lab, which he co-founded in late 2017.
transformation advisory practice. Open Innovation & Networks, Trend Scouting & Fore-
sighting, as well as the Development of the Corporate
Philipp Gerbert is Future Shaper of UnternehmerTUM AI Strategy are key areas which he influenced within
and Director at appliedAI. Previous activities include Siemens in recent years and still thrives to drive them
many years as Senior Partner and Lead of Digital further. He holds a diploma in “international cultural and
Strategy at BCG, Fellow for AI in Business at the business studies” and an M.A. for “musical education”
BCG Henderson Institute, as well as a Partner at the from the University of Passau, and started his career in
McKenna Group in Silicon Valley. He holds a PhD in 2006 at Siemens Management Consulting.
Physics from the Massachusetts Institute of Tech-
nology (MIT). Vera Eger is part of the Siemens AI Lab Family since
November 2019. She studied Psychology in Regensburg
Philipp Hartmann serves appliedAI as Director of AI and is doing her master’s in Economic-, Organizational-
Strategy. Prior to joining appliedAI, he spent four years and Applied Social Psychology at the LMU. Vera is parti-
at McKinsey & Company as a strategy consultant. cularly interested in research about human-machine
Philipp holds a PhD from Technical University of interactions and the resulting challenges for society
Munich where he investigated factors of competitive and economy.
advantage in Artificial Intelligence.

Andreas Liebl is Managing Director at UnternehmerTUM

as well as appliedAI. Before joining UnternehmerTUM,
he worked for McKinsey & Company for five years and
completed his PhD at the Entrepreneurship Research
Institute at the Technical University of Munich.

Maria Schamberger is Senior AI Strategist at appliedAI.

She has a rich background within the financial services
industry from her former role as Vice President at
the Allianz Group as well as consulting and research
experience from McKinsey & Company. Maria
studied Corporate Innovation at Stanford University
and Banking at the Frankfurt School of Finance and
Management.

Alexander Waldmann is the Director of Operations

at appliedAI. Before joining appliedAI he served as
Visionary Lead of [x], UnternehmerTUM‘s unit for
disruptive technologies and helped create ~300
startup-teams during that time. Alex studied Computer
Science at the Technical University of Munich and MIT.

The authors would like to thank Annette Bauer for her

support in writing this report and Henrike Noack for
designing this publication.
About appliedAI
The appliedAI Initiative, Europe’s largest non-profit
initiative for the application of artificial intelligence
technology, aims to bring Germany into the AI age and
offers its wide ecosystem of established companies,
researchers, and startups neutral ground on which to
learn about AI, implement the technology, and connect
with each other. NVIDIA, Google, MunichRe, Siemens,
Deutsche Telekom, and many more are partners of the
initiative, which started in early 2018.

You can find more information about appliedAI at:

www.appliedai.de
UnternehmerTUM GmbH
Lichtenbergstraße 6
D-85748 Garching
Deutschland