Proceedings of the 57th Hawaii International Conference on System Sciences | 2024

A Framework of AI Strategy

Yao Shi Judith Gebauer Elahe Javadi

University of North Carolina University of North Carolina Illinois State University
Wilmington Wilmington ejavadi@[Link]
shiy@[Link] gebauerj@[Link]

Abstract AI-assisted tasks and hybrid decisions results in changes

Building a vibrant strategy for the use of Artificial to business models, organizational structures, and, in
Intelligence (AI) in business is becoming critical for turn, changes to the value propositions of organizations
company success in the AI era. To decode AI strategy, (Dwivedi et al., 2021). Pioneer organizations in private
this study aims to address three questions: (1) What is and public sectors and even national governments have
AI strategy? (2) What are the key elements of AI started to develop strategies for adopting, advancing,
strategy? and (3) How can AI strategy be applied and and applying AI technology to create opportunities and
impact a business? We define AI strategy and propose value throughout their business activities (Joshi &
an AI strategy framework consisting of three layers: AI Wade, 2020).
strategic competency, AI use cases, and AI enabling Although the importance of AI has been well
factors. Based on published cases that describe early recognized, a great number of studies focus on technical
applications of AI, we identify a bottom-up approach questions about developing or applying AI but there is a
typically used by Small and Medium-sized Enterprises lack of studies that explore AI with respect to strategic
(SMEs) in building AI strategy and a top-down management (Berente et al., 2021; Keding, 2021). AI
approach typically used by big tech firms and applications reflect the heterogeneity of the scenarios in
traditional incumbent firms. We also find differences in which they are utilized and have started to penetrate
the core of the AI strategies that are initiated by the many facets of business operations. However, common
three types of enterprises: SMEs: survival oriented; big guidelines for strategy building are still missing. With
tech firms: AI ecosystem oriented; traditional this study, we intend to move closer to providing an
incumbent firms: business alignment oriented. overarching and systematic guide for enterprises
seeking to build a vibrant AI strategy. Specifically, we
Keywords: AI, AI Strategy, Competitive Advantage, focus on three questions: (1) What is AI strategy? (2)
Digital Transformation. What are the key elements of AI strategy? and (3) How
can AI strategy be applied and impact the business?
1. Introduction To answer those questions, we review previous
studies that apply strategic views of technology and that
can be applied to AI. IS strategy has been defined as “the
Artificial Intelligence (AI) is becoming a critical
organizational perspective on the investment in,
element for the competitiveness, sustainable
deployment, use, and management of information
development, and overall success of companies because
systems” (Chen et al., 2010, p. 237). Applying Chen et
of its incredible recent advances and ubiquitous
al.’s (2010) notion to AI, we define AI strategy as the
applicability in business and society (Di Vaio et al.,
organizational perspective on the investment in,
2020; Krakowski et al., 2023; Mahidhar & Davenport,
deployment, use, and management of AI technology and
2018). AI is used to enhance the experience of
AI systems. In addition, a shared perspective
customers by intelligently processing vast amounts of
(Mintzberg, 1987) on AI strategy and its components is
customer data (Huang & Rust, 2022), to manage
necessary to guide the selection of development and
repetitive and routine knowledge work tasks (Davenport
adoption alternatives, the acquisition and preparation of
& Ronanki, 2018), to leverage strategy and optimize the
data sources (Correani et al., 2020), training and
measurement of key performance indicators in
reconfiguration of models, and the maintenance and
organizations (Kiron & Schrage, 2019), and to handle
fine-tuning of AI infrastructure and services. A shared
many more tasks that have not been explored yet. The
perspective of AI refers to a common view of a given
transition from human-oriented tasks and decisions to

URI: [Link]
978-0-9981331-7-1 Page 227
(CC BY-NC-ND 4.0)
organizational locus and direction in a particular area From a practical perspective, strategy can be
which is communicated clearly with all relevant developed not only for an entire organization, but also
stakeholders. Nowadays, the development of AI for any of its units, sub-units, and functions across the
pipelines is time consuming, meaning that wait-and-see organizational hierarchy, including information systems
approaches can be problematic in a fast-paced (IS) as a whole or for parts of it. Maintaining
environment (Mahidhar & Davenport, 2018). consistency among the various strategic plans is
Formulating an AI strategy enables a learn-fast and fail- important to avoid conflicts and inefficiencies.
fast approach instead. Building on prior literature, we At the organizational level, Chen et al. (2010)
refine our notion of AI strategy with a focus on both describe the following three concepts of IS strategy: “(1)
existing AI operations as well as future initiations. IS strategy as the use of IS to support business strategy;
Our study is structured as follows: We start with a (2) IS strategy as the master plan of the IS function; and
literature review in section 2, then present a framework (3) IS strategy as the shared view of the IS role within
for AI strategy that integrates the core elements the organization (p. 238, emphases in the text). The
identified in the literature review (section 3). The third conception, IS strategy as a shared view of the IS
framework combines elements in three layers. The top role within the organization is suggested to be most
layer, AI strategic competency, includes AI vision, appropriate to be applied in the context of building a
competitive advantage, and customer value. The middle strategy for impactful innovations because it best
layer, AI use cases, includes AI-powered products and balances the perspectives of the business and IS related
services and AI-powered operations. The bottom layer, to the technology (Chen et al., 2010).
AI enabling factors, includes people, organization, Correani et al. (2020) emphasize the need to
technology, and AI ecosystem. In section 4, we examine properly execute a strategy to achieve its value.
published case studies of AI applications and identify Following the definition of a “guiding policy for the
two approaches for building an AI strategy: the bottom- creation and appropriation of value by exploiting digital
up approach and the top-down approach. While the technologies to achieve long-term objectives” (i.e.,
former method is used primarily by Small and Medium- strategy formulation), firms need to “translate the digital
sized Enterprises (SMEs), the latter method is used more strategy formulated into a concrete plan and set of
often by big tech firms and traditional incumbent firms. actions” (p. 38). To achieve consistency between the
In addition, we identify differences in the core of the AI actions and objectives, Correani et al. (2020) highlight
strategies that are initiated by the three types of the four building blocks of resources, capabilities,
enterprises. SMEs’ AI strategies are survival-oriented, activities, and stakeholders and emphasize shifts that
big tech firms’ AI strategies are AI ecosystem-oriented, may happen in relation to technology innovations. For
and traditional incumbent firms’ AI strategies are example, the role of data might be more central in the
business alignment-oriented. Section 5 concludes the recent context of digital transformation and AI than
paper. more conventional business models of the past.

2. Literature Review 2.2. AI and AI Strategy

2.1. Strategy and IS Strategy In the following, we first point out previous work
on the particularities of modern AI, before we review
Strategy refers to a long-term plan for how a studies that have suggested the need for comprehensive
business is going to compete, what its goals should be, and long-term AI strategies and that have provided
and what activities the company will perform to carry starting points for AI strategy development.
out those goals (Mintzberg, 1987). Ideally in line with a AI is commonly associated with the capability of
higher-level and stable vision for the organization machines to emulate certain traits of human intelligence
(Collins & Porras, 1996), a good strategy can allow a (Benbya et al., 2021; Russell & Norvig, 2020). While
company to outperform its rivals by delivering a unique AI has been around for many decades, recent advances
mix of values (Porter, 1996). As a strategy is in computing power and algorithms, as well as the
implemented, it informs decisions about resource availability of large volumes of high-quality data and
allocation, operations, and specific projects. While powerful, typically cloud-based, storage solutions have
strategies have long been implemented in a somewhat accelerated its commercialization and establishment in
static fashion to cover a period of about three to five business operations and everyday lives (Benbya et al.,
years, the fast-paced developments of contemporary 2021; Dwivedi et al, 2021).
business environments have resulted in more dynamic, Berente et al. (2021) emphasize the dynamic aspect
rolling approaches to strategy (Teece et al., 2020). of AI as a moving target when they refer to it as “the
frontier of computational advancements that references

Page 228
human intelligence in addressing ever more complex Mahidhar and Davenport (2018) point out that
decision-making problems” (p. 1435). Contemporary adopting and fully realizing the value of AI can take
forms of AI are characterized by the three interrelated longer than for regular IT, thus requiring more upfront
facets of autonomy, learning, and inscrutability (Berente planning (i.e., strategizing) for several reasons. AI
et al., 2021). Autonomy means that AI-systems can act systems often need to be configured to fit an
on their own and make autonomous decisions, including organization’s unique situation. Then they also need to
generative behavior and new ways of interacting with be trained with data that may first need to be collected
human and artificial agents, all with possibly profound and prepared. For areas where knowledge domains have
implications for management and knowledge work. not yet been modeled by a vendor, a knowledge
Learning refers to the ability of AI-systems to improve architecture may have to be set up and configured. AI
through data and experience inductively and must be integrated into an organization’s processes. And
automatically. Learning includes new methods of deep it needs to be decided what implications the AI systems
and reinforcement learning that utilize ever more have for the workforce. All these aspects increase the
powerful computing and new forms and very large urgency of developing an AI-related strategy sooner,
quantities of data within and beyond the confines of the rather than later.
organization. AI-powered systems can be capable of Joshi and Wade (2020) indicate that many
generating new knowledge with limited human companies use AI mainly for “tactical rather than
oversight or input. Inscrutability refers to the use of strategic purposes” and a “cohesive long-term AI
algorithmic models that are intelligible to at most very strategic vision is rare” (Joshi & Wade, 2020). To
few humans, characterized by varying levels of opacity, address the lack of longer-term strategic vision and
transparency, explainability, and interpretability with support the development of a sustainable approach to
implications for the “black box” phenomenon often AI, Joshi and Wade (2020) suggest building an AI
attributed to AI and other management issues. strategy on three pillars: (1) robust and reliable
Among the different forms of contemporary AI technology infrastructure, in particular as it relates to
(e.g., robotics, expert systems, neural networks), usable data and feedback loops; (2) new AI-enabled
machine learning (ML) has been labeled as the most business models; and (3) AI-related ethics to counter
relevant and lucrative (Benbya et al., 2021; Hartman, AI-related bias and ensure accountability.
2023; Jacobides, 2021). ML “uses statistics to find Li et al. (2021) define AI orientation, which
patterns in large amounts of data and uses these patterns corresponds with our notion of AI strategy, as “a firm’s
to make predictions” (Hartman, 2023, p. 2). Recent overall strategic direction and goals associated with
applications of AI can provide recommendations and introducing and applying AI technology,” which
increasingly automated support for complex decisions “provides managers with a framework to guide how
in areas, such as knowledge work in banking, insurance, they invest, manage, and apply AI within their firms'' (p.
customer relationship management, drug discovery, 1604). The underlying goal of developing an AI
toxicology, or bioinformatics (Benbya et al., 2021). The orientation is to “maximize the benefits of AI while
discussion about to what extent and how AI substitutes minimizing the associated costs and risks'' (p. 1604). It
(automates) or complements (augments) human is intended to help managers make decisions regarding
intelligence is ongoing (Krakowski et al., 2023). investments, applications, and communication with
Important components of modern AI systems relate stakeholders. The need for AI-related strategy results
to data management (sourcing and integration of large from the specific characteristics of the technology but
volumes of high-quality data), system and model also from its novelty and the many associated
development and refinement, and deployment unknowns. Aspects to be included in strategies related
infrastructure, including computation and storage to current developments of AI include not only the
(Benbya et al., 2021; Jacobides et al., 2021). provisioning of adequate technology and systems that
Interestingly, the developers and providers of critical AI are sufficiently powerful and scalable, but also
components, such as Google, Amazon, Microsoft, and objectives, guidelines, policies, and standards to assess
other members of “big tech”, are also among AI’s the potential for AI-based and -supported applications
biggest users and benefactors (Jacobides et al., 2021). and their risks and limitations, determine specific uses,
We also note that AI is conceptually close to digital evaluate the outcomes of AI-based decisions and AI-
transformation because it enables and supports the use generated content, and manage the implications of AI in
and interpretation of large amounts of data that drive general.
digital transformation (Gurbaxani & Dunkle, 2019). Kiron and Schrage (2019) suggest using key
Consequently, AI can play a role in taking digital performance indicators (KPIs) to define a company’s AI
transformation to the next level (Correani et al., 2020). strategy that then provides corporate leaders with a basis
to create value, accountability, and competitive

Page 229
advantage. They also emphasize the need to develop a relevance, and quality, we limited the search results to
strategy not just for AI but also with AI and state that the articles published in top journals in IS (Decision
“optimizing carefully selected KPIs becomes AI’s Support Systems, European Journal of Information
strategic purpose” (p. 31). For example, ML could be Systems, Information & Management, Information and
applied to choose, measure, and optimize the Organization, Information Systems Journal,
organization’s selected KPIs, whereby data and careful Information Systems Research, Journal of Information
data management again play an important role to Technology, Journal of Management Information
enhance the measures and help the machines learn over Systems, Journal of Strategic Information Systems,
time. Journal of the Association of Information Systems, MIS
Krakowski et al. (2023) argue that AI as a potential Quarterly) and management (Academy of Management
supplement and complement of human intelligence Journal, Academy of Management Review, Decision
alters the sources of competitive advantage. Managers Sciences, Harvard Business Review, Management
have “to develop new capabilities to stay relevant in an Science, Strategic Management Journal). The database
AI-based competitive landscape” (p. 1426). AI can search yielded 431 articles. In addition, we applied the
substitute domain-specific capabilities, e.g., related to same search teams and relevant topics in Google
managerial decision making, if it helps non-experts Scholar. After reviewing the titles, abstracts, and full-
perform better and consequently erodes the competitive texts of the articles, we excluded duplicates and articles
advantage of the domain experts. At the same time, there of no and low-relevancy to AI strategy. The screening
can also be a complementary effect when the technology process yielded 29 articles (12 from EBSCOhost; 17
(which is general-purpose to some extent) is used to from google scholar) as the final relevant set of studies.
develop unique bundles of competencies that augment Based on the identified relevant studies (e.g.,
the domain knowledge. Competitive advantage can Berente et al., 2021; Bharadwaj, 2000; Floridi et al.,
consequently shift to those who best understand what to 2018; Hartman et al., 2023; Huang & Rust, 2022;
do with and how to apply the technology, which might Jacobides et al., 2021; Joshi & Wade, 2020; Keding,
be associated more with capabilities related to the 2021; Kemp, 2023; Krakowski et al., 2023; Raisch &
technology (e.g., AI literacy) or creativity than the Krakowski, 2021; Ravichandran, 2005), we propose an
original domain. AI strategy framework that consists three layers, each
A good AI-related strategy will understand the comprising of several elements: AI strategic
potential of AI to substitute traditional capabilities, e.g., competency, AI use cases, and AI enabling factors,
to make managerial decisions and solve business (Figure 1).
problems and ensure that the organization stays in play
as the technology is increasingly becoming a business 3.1. Top Layer: AI Strategic Competency
necessity. At the same time, a good AI-related strategy
should also consider ways to augment human expertise 3.1.1. AI Vision. AI vision constitutes the top layer of
with the goal of developing unique capabilities that have the framework. On the one hand, AI vision must be
the potential to afford new competitive advantages. under the arch of the business strategy which is not
included in the framework to align with the big picture
3. AI Strategy Framework of the company. On the other hand, AI vision directly
and indirectly reconciles all the elements in the
Under the umbrella of the strategy definition, it is framework. At the top layer, two important aspects
suggested that AI strategy is a part of a company’s derived from AI vision are competitive advantage and
business strategy and refers to a plan of action for how customer value (Collins & Porras, 1996). The
an organization intends to strategically build and stakeholders of AI strategy need to figure out: (1) What
leverage AI technologies to achieve its goals and is the situation of the company in the competition? (2)
objectives (Hartman et al., 2023). A well-defined AI Who are the existing and potential competitors before
strategy can help an organization align its AI initiatives and after adopting AI as part of the company’s business?
with the overall business strategy and to prioritize and (3) What is the real value that the company can provide
allocate resources to the most promising AI initiatives. to customers through AI?
To build a conceptual framework, we conducted a
literature review with the commonly accepted
guidelines (Rowe, 2014; Webster & Watson, 2002).
First, we performed a literature search in the
EBSCOhost database (Business Source Ultimate) with
the search terms of “AI” and “artificial intelligence” in
combination with “strategy”. For the sake of timeliness,

Page 230
 Figure 1. AI strategy framework

products, services, or experiences, or both. At this point,

3.1.2. Competitive Advantage. Gaining competitive it is important for AI strategy makers to define their
advantage through AI needs to deal with three strategic position with respect to customer value. The typical
limitations of AI: the generic, explicit, and myopic formats of AI applications are depicted in the middle
nature of AI, which have critical adverse effects (Kemp, layers.
2023). AI is built primarily on algorithms, data, IT
infrastructure, and computing devices (Jacobides, 3.2. Middle Layer: AI Use Cases
2021). Since hardware like IT infrastructure and
computing devices can be easily acquired in the market, Linked to the top layer, the AI use case layer defines
algorithms and data become the critical factors to create how to provide value to users. As AI enables machines
unique capabilities. However, many algorithms for and systems to perform tasks like a human and to
building AI features such as natural language processing become more intelligent, the range of applications and
are generic, which can mean similar results for use cases increases. The strategic value of AI manifests
companies using the same algorithms. For example, “all itself in the form of products, services, and operations
else being equal, a neural network algorithm will arrive that benefit from AI at some level as AI provides
at the same logic for connecting a set of inputs to outputs augmentation or automation or both (Raisch &
regardless of whether Spotify or Pandora Krakowski, 2021). In general, we classify AI use cases
operationalizes the neural network” (Kemp, 2023, p. 7). into two categories: AI-powered product and service,
In addition, the basic element of algorithms is code and AI-powered operation.
which is explicit and can be replicated by competitors.
Last, AI is good at handling tasks given a specific scope 3.2.1. AI-powered Product and Service. Many
but lacks contextual awareness without a clear boundary organizations have designed new products and services
such as a firm’s strategy. Therefore, designing unique or improved already existing ones through the power of
AI use cases in the middle layer, capturing unique data, AI to gain competitive advantage and deliver value for
and synthesizing the internal technologies as well as their customers (Kemp, 2023). We are seeing many
other resources at the bottom layer for training AI are cases across industries. For instance, AI-enabled
critical steps to building competitive advantages. chatbots can provide 24/7 customer service, handle
frequently asked questions, and support customers in
3.1.3. Customer Value. AI techniques have been used real-time (Davenport, 2018). ByteDance which
to provide value to customers in many areas. Although developed TikTok has gained significant competitive
the resulting product or service value does not always advantage among social platforms embedded with AI
signal the use of AI, the technology can provide a through providing value to the content creators through
smooth, easy, or seamless user experience (Davenport, advanced content-recommendation (Davis et al., 2019).
2018; Di Vaio et al., 2020; Huang & Rust, 2022). The C3 develops products and services for the Air Force that
customer value created through AI typically results leverage services provided by cloud AI platforms to
either from improving work efficiency and in
consequence lowering cost or from creating unique

Page 231
create systems for preventative maintenance of questions about the impact on organizational business
mechanical systems (Simon & Gonzalez, 2018). processes. In some cases, there are change management
and internal conflicts related to the adoption and
3.2.2. AI-powered Operation. The management implementation of AI. For example, PittaRosso, the
literature has identified two broad areas of AI Italian shoe retailer, faced internal conflicts between
applications: (1) automation applications where the finance and sales when it adopted and implemented an
tasks and processes are completed by the machines only; AI-based pricing and promotion system (Israel, 2021).
(2) augmentation where tasks and processes completed To orchestrate the interaction with other enabling
by humans are improved by machines (Brynjolfsson & factors and the upper layers of the framework, the
McAfee, 2014; Davenport & Kirby, 2016; Dougherty & stakeholders of AI strategy need to address questions
Wilson, 2018). However, Raisch & Krakowski (2021), related to organizational structure and governance, such
argue that although the automation-augmentation as: what is the key support the AI project can expect
distinction is intuitive, it overlooks the from the organizational change? which
interdependencies among the two categories of department/business units will be impacted by the AI
applications and that their boundaries are blurred across project? how will AI change the culture in the company?
time and space. The augmentation versus automation is how should the change process be managed?
sometimes also termed complementation versus
substitution through AI (Krakowski et al., 2023) and 3.3.3. Technology. Technology constitutes another key
may be a valid starting point for the formulation of an enabling factor. The IS infrastructure and data are two
AI strategy. FireEye, for instance, applies AI to detect primary facets of technology in the framework. AI relies
suspicious network activities but also relies on human heavily not only on powerful computational devices and
experts to identify new potential cyber-attack scenarios huge amounts of data needed to train AI models, but also
(Miller & Bhattacharya, 2021). on reliable and fast networks that transmit data among
numerous digital devices. Therefore, stakeholders of AI
3.3. Bottom Layer: AI Enabling Factors strategy need to understand available technology
resources in the organization. More importantly, they
AI enabling factors constitute the bottom layer in must be clear on what technology resources are lacking
the framework and provide the foundation for AI that hinder realizing the AI project.
strategy and implementation. Although many factors
may impact the success of an AI project, people, 3.3.4. AI Ecosystem. The AI ecosystem interconnects
organization, technology, and the AI ecosystem are other enabling factors in the framework. Most
identified as indispensable elements supporting companies do not have sufficient resources in the form
company growth in many studies (Bharadwaj, 2000; of people, data, or infrastructure to develop an AI
Hartmann et al., 2023; Jacobides et al., 2021; project on their own. An AI ecosystem can enable
Ravichandran et al., 2005). industries where companies want to collaborate with
partners who possess the resources or techniques that
3.3.1. People. People constitute a critical part of any they need (Jacobides et al., 2021). This channel can ease
project and the related changes. Organizational human the challenges associated with two critical AI enabling
resources significantly impact the progress and result of factors: people and technology.
an AI project. People include both new management Jacobides et al. (2022) identified three distinct areas
roles that drive AI projects and the employees who in an AI ecosystem: AI enablement, AI production, and
possess the skills and capabilities associated with AI consumptions. The ecosystem involves developers,
developing and applying or interacting with AI. manufactures, and users of AI. Depending on the core
Companies planning to step into the AI area report business of an organization and the AI use cases, any or
challenges related to the review of in-house AI skills and a combination of these may characterize the role of the
capabilities, training current employees, and hiring and organization in the ecosystem. In the context of AI
retaining AI experts. Furthermore, establishing and strategy within a single firm, the AI ecosystem refers
fostering trust between human operators and the AI those factors that motivate, enable, mandate, or make
tools they interact with is essential. feasible the aspiration of the organizations for AI-
related innovations in products, services, and
3.3.2. Organization. AI is a powerful technology that operations. The ecosystem involves entities such as
has the potential to change organizational structures and hardware manufacturers like NVIDIA (Watkins et al.,
governance in many ways and transform individuals’ 2019) and platform providers like AWS or Numenta that
positions in organizations. The changes can create enable AI products provided by downstream
concerns for employees that would lead them to raise organizations such as C3 (Simon & Gonzalez, 2018) as

Page 232
well as suppliers, partners, or regulatory bodies that industry background notes that we retrieved from the
demand an advancement on the augmentation or Harvard Business Publishing Education database
automation scale of the company. ([Link] using the keyword “AI
Strategy”. The examined cases demonstrate a variety of
3.4. AI Governance strategic roles that AI plays as it is adopted in business
and the implications it has for business transformation.
As AI, to some extent, has an ability of sensing the Based on the cases that we reviewed, we find the AI
environment and is being trained to think and behave strategy framework to be applicable, as AI-related
like a human, a concern might arise in cases where the strategies are described explicitly or implicitly in the
output generated by an AI system cannot be fully papers. However, we also noticed that the AI strategy
controlled (Berente et al., 2021) and can be associated framework is not a one-size-fits-all framework. It plays
with all the elements from the three layers. Therefore, out differently in different types of companies to
we suggest regarding AI governance as a more general accommodate complex situations.
component that transcends the other layers.
Many test results of AI applications might be 4.1. Cases in Small and Medium-sized
considered misleading, biased, and even dangerous. Enterprises (SMEs)
They are not ethical or adhere to regulation, law, and
convention (Benjamin, 2019; Jobin, 2019). For instance, Small and midsize enterprises (SMEs) are referred
ChatGPT is a chatbot which interacts in a conversational to as businesses with revenues, assets, or number of
way and has quickly gained attention for its detailed employees below a certain threshold. According to the
responses and articulate answers across many domains standards of U.S. Small Business Administration (2023)
of knowledge. However, its uneven factual accuracy and considering that most AI companies fall into the
was identified as a significant drawback. People have information industry, we classify SMEs as companies
used ChatGPT to flood Stack Overflow, the go-to with less than 1,500 employees.
question-and-answer site for coders and programmers, SMEs, including startups, usually initiate AI
with AI responses that seem correct at first glance but strategy from the enabling factors and the corresponding
are often wrong on close examination. The site has AI use cases in the AI strategy framework. As the AI
temporarily banned users from sharing responses technology becomes mature and can be applied to a
generated by ChatGPT. larger scale of customers and users, the higher level
Related to AI applications in management, elements of the AI strategy framework, such as AI
assessing the consequences, evaluation equity, privacy, vision (e.g., what is the company’s vision on adopting
and transparency (Castelvecchi, 2016) of impacts are AI in its business?) and AI governance (e.g., how to deal
pressing issues. Only a fraction of organizations has with potential ethical issues or conflicts in AI
started to explicitly address ethical governance. application?), must be taken into greater consideration
Mastercard, for instance, has leveraged its existing to help ensure that a vibrant AI strategy can guide the
privacy framework to include assessment, mitigation, AI application as part of the business. We call this
and auditing of AI-related processes (Isik & Duke, approach to building an AI strategy Bottom-Up
2022). Initiation (BUI).
AI governance plays an important role across the Most BUI occurs in SMEs because they are small
layers of our AI strategy framework because the and possess fewer resources that are crucial for survival
underlying issues are complex and often rooted in than large companies. Especially in startups, the core
multiple perspectives. Key questions to ask include: assets tend to be the founders and employees who are
How can we avoid invading people’s privacy? Are there experts in certain areas of AI. SMEs often provide
any legal implications of using AI in this way? What unique AI products or services (e.g., ChatGPT; vision
sort of consent do we need from recognition for self-driving vehicles). They focus on
customers/users/employees? How can we ensure our AI niche markets that are less likely covered by incumbent
is free of bias and discrimination? or large size companies. Given the characteristics of
SMEs and relatively lower competition, this type of
4. AI Strategy Framework Applications companies tends to start forming a systematic AI
strategy from the bottom of the framework (i.e., AI
The AI strategy framework is intended as a guide enabling factors). Those crucial resources determine
for companies who plan to or are on the trajectory to how the SMEs serve their customers and directly impact
adopt AI in their business. To assess the applicability of the survival in the market. In contrast, the higher-level
the framework, we examined 65 cases out of 417 elements of the framework, i.e., AI vision, competitive
searched articles, book chapters, case studies, and advantage, customer value, as well as AI governance,

Page 233
tend to not directly influence survival in the short term. built models to be used for organization-specific
Therefore, the higher-level perspectives of AI strategy services. By providing customer value to other
are often built and reshaped at later stages. Therefore, companies, these large players are also trying to stay
we call this type of AI strategy “survival oriented.” competitive and capture market shares. Hence, some of
the cases we reviewed reflect more than one area in the
4.2. Cases in Large Enterprises AI strategy framework. In the context of developing AI-
based competencies, organizations provide programs
To distinguish large enterprises from SMEs, we for intelligent marketing, human-resource development,
classify large enterprises as companies with more than manufacturing, and business operations management.
about 1,500 employees. In contrast to the BUI approach Intel Corporation’s AI for Youth (Unruh, 2021)
that is typical for SMEs, large enterprises often start exemplifies a multinational organization investing in AI
building AI strategies from the top of the AI strategy for human resource development to ensure that
framework (i.e., AI vision, competitive advantage, manufacturing and operations needs are met with local
customer value) and define the direction of AI adoption talents.
at a higher level for the organizations. During the Traditional incumbent firms are large enterprises
strategy making process, large enterprises tend to be with primary business activities outside of the
more rigorous and cautious in shifting the direction of information and high technology sectors. They usually
their mature businesses and adopting AI into their have dominant positions in traditional industries, such
organizations. Like the waterfall method in software as retail, accommodation, and finance. Traditional
development, this methodical, step-by-step approach incumbent firms tend to lack specific experience and
tries to anticipate and minimize mistakes that may resources for developing and applying AI. The primary
jeopardize existing businesses and operations. We label concern related to the utilization of AI is to explore new
this approach to building an AI strategy Top-Down opportunities that can serve the current business.
Initiation (TDI). Therefore, for incumbent firms, the process of AI
Based on our review of published AI-related case strategy development should begin at the top of the
studies, we identified two types of large enterprises: big framework. It is critical to align the AI vision with the
tech firms and traditional incumbent firms. Both possess current business vision of the company (Keding, 2021),
mature resources and customers, are dominant in their and use the vision to link and guide the components in
industries, and tend to have comparatively mature the middle and bottom layers of the framework.
business strategies, products, processes, organizational Questions to be addressed include: How to embed AI
structures, and personnel. In the short term, they don’t technology into current products? How to redesign
have survival threats. While TDI appears to be the business processes and revenue models to accommodate
method used by both types of big enterprises in building AI? and How to allocate current resources for AI
their AI strategy, we also found that big tech firms tend development and application? Examples of this type of
to focus more on maximizing value from building an AI effort are the Hotel Innovation Incubator at Charlotte
ecosystem (“AI ecosystem oriented”), whereas Marriott City called M Beta intended for identifying and
traditional incumbent firms tend to focus more on testing strategically aligned customer-facing
maximizing value from aligning AI to existing innovations and the Global Creative Intelligence at
businesses (“business alignment oriented”). McCormick and Company (Meister & Chandrasekhar,
Big tech firms are large enterprises with primary 2021). Allianz used data and machine learning to
business activities in the information and high alleviate plummeting sales and predict direct debits
technology industry, such as Google, Amazon, and (Heyvaert et al., 2021; Van der Shraelen et al., 2022).
Microsoft. They are the key participants in the AI Investment firm Vanguard has started a new “Personal
ecosystem, as they control or have penetrated most or Advisor Service”, which combines AI technology with
all stacks of AI enablement (e.g., data investment advising that was traditionally provided
management/processing, computing, infrastructure) and primarily by human advisors. In the redesigned service,
AI production (e.g., machine learning/data analysis AI replaces human advisors to perform the traditional
platform, visualization) (Jacobides et al., 2021). Relying advising tasks such as customizing investment
on the foundations, which requires significant and long- portfolios. Human advisors play the role of coaches to
term investments, the big tech firms not only create AI answer customer questions and keep customers on the
for internal use but also develop products, services, and right track. This redesigned service stands out from
technologies that offer AI affordances to other purely human advising services. It has significantly
companies. Google’s AutoML, for instance, allows reduced the cost for Vanguard and helped customers
customer organizations to employ Google services for increase their assets (Davenport & Ronanki, 2018).
underlying data acquisition and management, and pre-

Page 234
5. Implications and Limitations Collins, J. C., & Porras, J. I. (1996). Building your company's
vision. Harvard Business Review, 74(5), 65-77.
Correani, A., De Massis, A., Frattini, F., Petruzelli, A. M., &
In this paper, we define AI strategy and propose an Natalicchio, A. (2020). Implementing a digital strategy:
AI strategy framework to enhance the theoretical Learning from the experience of three digital
foundation of AI governance. To provide practical transformation projects. California Management Review,
implications for top executives, AI professionals, and 62(4), 37–56.
researchers in the area, we analyze published case Daugherty, P. R., & Wilson, H. J. (2018). Human+ machine:
studies of AI strategy applications. The framework Reimagining work in the age of AI. Harvard Business
provides a tool for organizations to identify and describe Press.
their current state with respect to the core competencies, Davenport, T. H. (2018). The AI advantage: How to put the
artificial intelligence revolution to work. Cambridge: The
resources, and organizational vision for the future. This
MIT Press.
understanding of the current state of AI can help Davenport, T. H., & Kirby, J. (2016). Only humans need
articulate desired future states and potential required apply: Winners and losers in the age of smart machines.
changes. New York: Harper Business.
We acknowledge limitations of our study. For Davenport, T. H., & Ronanki, R. (2018). Artificial intelligence
example, we have relied on published articles and case for the real world. Harvard Business Review, 96, 108–
studies to build the framework and explore the patterns. 116.
The descriptions found in the cases may not fully reflect Di Vaio, A., Palladino, R., Hassan, R., & Escobar, O. (2020).
the whole picture of AI strategy across companies and Artificial intelligence and business models in the
sustainable development goals perspective: A systematic
industries. Future research is needed to explore the topic
literature review. Journal of Business Research, 121,
further, including quantitative research and natural 283-314.
experiments to examine the impact of the AI framework Dwivedi, Y. K., Hughes, L., Ismagilova, E., Aarts, G.,
and to develop a deeper understanding about the Coombs, C., Crick, T., ... & Williams, M. D. (2021).
causality between its various components. Artificial Intelligence (AI): Multidisciplinary
perspectives on emerging challenges, opportunities, and
6. References agenda for research, practice and policy. International
Journal of Information Management, 57, 101994.
Floridi, L., Cowls, J., Beltrametti, M., Chatila, R., Chazerand,
Benbya, H., Pachidi, S., & Jarvenpaa, S. (2021). Special issue P., Dignum, V., Luetge, C., Madelin, R., Pagallo, U.,
editorial: Artificial intelligence in organizations: Rossi, F., Schafer, B., Valcke, P., & Vayena, E. (2018).
implications for information systems research. Journal of AI4People—An ethical framework for a good AI society:
the Association for Information Systems, 22(2), 281-303. Opportunities, risks, principles, and recommendations.
Benjamin, R. (2019). Assessing risk, automating racism. Minds and Machines, 28(4), 689-707.
Science, 366(6464), 421-422. Gregory, R. W., Henfridsson, O., Kaganer, E., & Kyriakou, H.
Berente, N., Gu, B., Recker, J., & Santhanam, R. (2021). (2021). The role of artificial intelligence and data
Managing artificial intelligence. MIS Quarterly, 45(3), network effects for creating user value. Academy of
1433-1450. Management Review, 46(3), 534-551.
Bharadwaj, A. S. (2000). A resource-based perspective on Gurbaxani, V., & Dunkle, D. (2019). Gearing up for
information technology capability and firm performance: successful digital transformation. MIS Quarterly
An empirical investigation. MIS Quarterly, 24(1), 169- Executive, 18(3), Article 6.
196. Hartmann, P., Liebl, A., Waldmann, A., & Brakemeier, H.
Brynjolfsson, E., & McAfee, A. (2014). The second machine (2023). Applying AI: The elements of a comprehensive AI
age: Work, progress, and prosperity in a time of brilliant strategy. Initiative for Applied Artificial Intelligence.
technologies. WW Norton & Company. [Link]
Burnap, A., Hauser, J. R., & Timoshenko, A. (2023). Product itepaper_Strategy_Screen.pdf.
aesthetic design: A machine learning augmentation. Heyvaert C. E., Darmawan V., Stouthuysen K., Verdonck T.,
Marketing Science. Grumiau C., & Thoppan Mohanchandralal, S. (2022).
[Link] Allianz: Improving P&L through machine learning.
Castelvecchi, D. (2016). Can we open the black box of AI? Vlerick Business School.
Nature, 538, 20–23. [Link]
Chen, D. Q., Mocker, M., Preston, D. S., & Teubner, A. 3.
(2010). Information systems strategy: Huang, M. H., & Rust, R. T. (2022). A framework for
reconceptualization, measurement, and implications. MIS collaborative artificial intelligence in marketing. Journal
quarterly, 34(2), 233-259. of Retailing, 98(2), 209-223.
Cohen, I. G., Evgeniou, T., Gerke, S., & Minssen, T. (2020). Iansiti, M., & Lakhani, K. R. (2020). Competing in the age of
The European artificial intelligence strategy: AI: Strategy and leadership when algorithms and
implications and challenges for digital health. The Lancet networks run the world. Harvard Business Press.
Digital Health, 2(7), e376-e379.

Page 235
Jacobides, M. G., Brusoni, S., & Candelon, F. (2021). The Simons R. & Golzalez, G. (2019). [Link]-Driven to Succeed.
evolutionary dynamics of the artificial intelligence Harvard Business School Publishing.
ecosystem. Strategy Science, 6(4), 412-435. Teece, D. J., Raspin, P. G., & Cox, D. R. (2020). Plotting
Jobin, A., Ienca, M., & Vayena, E. (2019). The global strategy in a dynamic world. MIT Sloan Management
landscape of AI ethics guidelines. Nature Machine Review, Fall, 28–33.
Intelligence, 1(9), 389-399. Unruh G. (2021). Intel Corporation: The AI for Youth
Joshi, A., & Wade, M. (2020, August 10). The building blocks Initiative. Thunderbird School of Global Management,
of an AI strategy. MIT Sloan Management Review. Harvard Business School Publishing.
[Link] U.S. Small Business Administration. (2023) Table of size
of-an-ai-strategy/. standards. U.S. Small Business Administration.
Isik, O., & Duke, O. (2022). Mastercard's ethical approach to [Link]
governing AI. Harvard Business School Publishing. standards
Keding, C. (2021). Understanding the interplay of artificial Van der Schraelen L., Willems E., Verdonck T., Grumiau C.,
intelligence and strategic management: Four decades of S. T., (2021). Allianz: Predicting Direct Debit with
research in review. Management Review Quarterly, Machine Learning. Vlerick Business School.
71(1), 91-134. Webster, J., & Watson, R. T. (2002). Analyzing the past to
Kemp, A. (2023). Competitive advantages through artificial prepare for the future: Writing a literature review.
intelligence: Toward a theory of situated AI. Academy of Management Information Systems Quarterly, 26(2), xiii–
Management Review, (In-Press). xxiii.
[Link]
Kiron, D., & Schrage, M. (2019). Strategy for and with AI.
MIT Sloan Management Review, 60(4), 30-35.
Krakowski, S., Luger, J., & Raisch, S. (2023). Artificial
intelligence and the changing sources of competitive
advantage. Strategic Management Journal, 44(6), 1425-
1452.
Li, J., Li, M., Wang, X., & Thatcher, J. B. (2021). Strategic
directions for AI: The role of CIOs and boards of
directors. MIS Quarterly, 45(3), 1603–1643.
Mahidhar, V., & Davenport, T. H. (2018). Why companies
that wait to adopt AI may never catch up. Harvard
Business Review, 12. [Link]
companies-that-wait-to-adopt-ai-may-never-catch-up.
Meister, D. & Chandrasekhar, R. (2021). McCormick & Co.:
Deploying Artificial Intelligence in New Product
Development. Ivy Business School Foundation.
Miller S. M. & Bhattarcharya, L. (2021). Cybersecurity at
FireEye: Human+AI. Singapore Management
University, SMU916.
Mintzberg, H. (1987). The strategy concept I: Five Ps for
strategy. California Management Review, 30(1), 11-24.
Porter, M. E. (1996). What is strategy? Harvard Business
Review (Nov-Dec), 61-78.
Raisch, S., & Krakowski, S. (2021). Artificial intelligence and
management: The automation–augmentation paradox.
Academy of Management Review, 46(1), 192-210.
Ravichandran, T., Lertwongsatien, C., & Lertwongsatien, C.
(2005). Effect of information systems resources and
capabilities on firm performance: A resource-based
perspective. Journal of Management Information
Systems, 21(4), 237-276.
Rowe, F. (2014). What literature review is not: Diversity,
boundaries and recommendations. European Journal of
Information Systems, 23(3), 241–255.
Russell, S., & Norvig, P. (2020). Artificial intelligence: A
modern approach. Prentice Hall.
Senoner, J., Netland, T., & Feuerriegel, S. (2022). Using
explainable artificial intelligence to improve process
quality: Evidence from semiconductor manufacturing.
Management Science, 68(8), 5704-5723.

Page 236