Elle Taurins, Faculty Lecturer, AIM Business School

With the recent launch of generative AI tools like ChatGPT, the discourse about AI in the corporate world ranges from its impact on job displacement and workplace productivity to risks and ethics. The expectations from AI are high worldwide: the market for AI is anticipated to reach 53 billion USD by 2026, with a compound annual growth rate of 35.4 during 2019-2026 (Barnea, 2020). At a C-suite level, the questions are posed on the AI capability to create a competitive advantage by formulating a strategic position or automating strategic decision-making (SDM). This article addresses the potential impact of AI on strategic thinking and strategy formulation. It overviews the benefits of AI tools currently in use and probes the extent to which SDM can be delegated to AI and ML (machine learning).

Strategising in the VUCA World

For leaders and managers, operating in an unpredictable and rapidly changing environment has become the ‘new normal’. The Volatile, Uncertain, Complex and Ambiguous business context, often referred to as VUCA World (Chadha, 2017), continues to intensify. The recent Covid challenge highlighted the need for strategic response capability in unpredictable situations. The unprecedented levels of complexity in both government and business contexts are emphasised by global and Australian researchers investigating the future of work (Ruppanner et al., 2023) and the business intelligence community (Barnea, 2020; Leavy, 2023).

Under VUCA conditions, organisations and corporations must be able to respond to pressures, identify new opportunities and make highly contextualised decisions. VUCA-responsive strategy formulation needs to be agile and less reliant on traditional tools and linear models. Thought leaders have been calling for new thinking and re-inventing strategy creation, suggesting impact analysis (Hines & Bishop, 2013), specific VUCA-factor response actions (Bennett & Lemoine, 2014), strategic foresight (Duijne & Bishop, 2018) and entrepreneurial approaches to strategy (Hoglund & Martensson, 2019).

Cognitive processes in strategy making

The concept of ‘strategy’ (or ‘strategy formulation’) and its adjacent notion of ‘strategic thinking’ are abstracts and are tricky to define. Mintzberg prominently stated (2000, p. 66) that “nowhere was anyone told how to create strategy. How to collect information, yes. How to evaluate strategy, yes. How to implement it, for sure. But not how to create it in the first place.” Strategic thinking underpins strategy formulation and SDM.

The construct of ‘strategic thinking’ is linked to advanced cognitive processes and a leader’s ability to think conceptually, however, it is continuously transforming and adapting to the needs of organisations. In the 1960s, it referred mostly to a capability to synthesise the data, in the 1990s, Mintzberg incorporated a manager’s ability to synthetise the learning in the direction of the business, and later the construct was enhanced into a “conceptual, systems-oriented, directional and opportunistic thinking leading to the discovery of novel, imaginative organizational strategies” (Casey & Goldman cited in Adzeh, 2017, p. 4). Strategic thinking frameworks like Liedtka’s five-factor ‘design approach’ model (1998) or a cognitive Pisapia et al.’s three-factor model (2005) imply that strategic thinking is both analytical and creative. It is meaningful only in a context, requires open minds, being comfortable with ambiguity and the ability to make consequential decisions (Pisapia, 2011).

In a VUCA world, a leader’s role requires a highly developed strategic thinking capability to make strategic decisions based on strategic options. Gartner Inc. has coined the term ‘Decision Intelligence’ (Larson, 2022) and argues, that Decision Intelligence is necessary for three types of inter-dependent decisions: one-off strategic decisions, repeated managerial decisions, and high-volume operational decisions. In essence, these types of decisions align with the three levels of strategy: corporate, business and functional (tactical).

In this turbulent business landscape and a rise of AI/ML, organisations pin their hopes on AI-generated strategic solutions. Many companies have a limited understanding of how the AI-driven change and disruption will impact their strategising and decision-making (Atsmon, 2023). Leaders still need to gain more clarity on the potential AI cognitive capabilities that could either meaningfully contribute to addressing corporate strategic issues, or substitute humans in SDM.

The key pain points in strategising processes

Many corporations are dedicating significant resources to gathering and analyse information on their competitors. Still, they often experience ‘strategic surprises’ when their competitors make moves that were not anticipated. Although the gap between the need for and the availability of the information is almost inevitable, the ‘fast yet reliable capture of emerging changes’ is critically important to senior executives driving strategy formulation (Barnea, 2020).

The traditional macro-environment analysis models like PESTLE require long lead times for data analysis because it relies on historical data and evidence-based reasoning. Considering that even Fortune 500 companies still use simple information-gathering tools, primarily from OSINT (open-source intelligence), the added value of such analysis to senior leaders needs to be improved (Barnea, 2020). This situation pushes executives to decide based on partial information or their ‘gut feeling’. The risk of bias in SDM is highlighted by many researchers and practitioners (Wu et al., 2023; Atsmon, 2023; Sanclemente, 2022; Hesel et al., 2022; Barnea 2020).

The speed of processing high volumes of data into manageable information and viable alternatives adds another pressure to SDM. In uncertainty, executives want to be as informed as possible. The data in organisations is often overwhelming and/or manually curated, and if the monitoring of the inputs is unsystematic or lagging, then important ‘weak signals’ may be overlooked or cause a prediction error. Incorrect predictions may be costly. Lagging data monitoring may lead to businesses not being able to respond in time and potentially losing their competitive advantage.

Strategy creation is a cognitively demanding process underpinned by cognitive insights; it involves reasoning, problem-solving and learning. To inform their decisions, strategists must identify underlying patterns, gain meaningful insights and use probability techniques to support their judgement.

In a nutshell, most organisations are experiencing the need for speed, greater anticipatory and predictive capability, and a synergy of analytical and judgement skills.

Where are we at with AI adoption in strategic applications?

AI is becoming embedded in our daily life, examples include Cortana, Alexa or Siri, or virtual assistants. Our business experience with AI is still rather limited, although task automation is growing in customer service and marketing. Empirical evidence shows that AI can be used to create marketing strategy, especially in manager-AI partnership strategy formulation activities (Eriksson et al., 2020). AI already performs functions such as analysing customer surveys or customer interactions, i.e. automated marketing decisions on an operational level (Barnea, 2020; Korzynski et al., 2023). Examples of AI use in marketing indicate that AI can indeed be used in strategy formulation, however, the strategy is limited to the functional rather than corporate level. Marketing uses a “Weak AI” type, i.e. it can emulate human logic by analysing large amounts of data and can act as a decision-maker if the decision process required is rational. Therefore, it can be automated and provide decision-making predictions or scenarios (options).

The 2022 study of 500 high-level B2C managers conducted by researchers from Nuremberg Institute for Market Decisions shows that 26% of these managers use AI in the role of “collaborator”, where humans interact with AI but control the overall process (Hesel et al., 2022). The preferred model among respondents was augmented decision-making with humans in control. Research suggested that humans are more receptive to AI when they can modify its decisions or forecasts, minimising the so-called ‘algorithm aversion’.

Although Hesel’s research team argues that the boundaries of AI in decision-making are shifting from the operational to the strategic level (Hesel et al., 2022), the definition of ‘strategic’ remains elusive and may not refer to the corporate level of strategy. Higher-order thinking skills would be needed for an organisation-wide strategy and SDM, or a ‘Creative Analytics’ type of AI to substitute a human in making a final decision related to imagination and creativity (Eriksson et al., 2020).

Key advantages of AI

AI undeniably exceeds human capabilities in translating big data, including unstructured inputs like audio, video, and images into manageable information and knowledge. This allows managers to input it into marketing and sales strategies. For example, when Netflix entered the content business, it made use of its data on 27 million US subscribers and 33 million subscribers worldwide. By using subscribers’ viewing history, searches, and ratings, Netflix decided to create a successful US adaptation of the British show House of Cards (Laney 2020 cited in Gudigantala et al., 2022)

Reduction of the long lead time for data analysis outputs may create a competitive advantage due to real-time data tracking (such as purchase data or search traffic) and speed of analysis (Barnea, 2020; Korzynski, et al., 2023).

AI algorithms can help to de-bias important executive decisions. AI can provide executives with competitor data analysis which will help to create a more objective understanding of their rivals by processing or even predicting competitors’ moves (Barnea, 2020).

The types of AI and why it matters

It is important to note that the outputs and benefits of AI may depend on the type of AI selected (Gudigantala et al., 2022; Wu et al., 2023). There is a plethora of AI definitions and classifications, and the discourse on AI/ML terminology is evolving in real-time.

In addition to an already mentioned “Weak” (task-specific) and ”Strong” (human-like problem solver), AI could be based on ability (narrow, general) or functionality (reactive, limited, self-aware), or it can be approached from a cognition (logical) or behavioural perspectives. Prior research organised AI systems into rule-based (decision process automation) or learning-based (predictive models using ML) and included an NLP (natural language-based system) to analyse customer reviews (Martinez-Plumed et al., 2021). AI could specialise in optimisation of solutions, or in physical interactions (robotics). Gartner Analytics Ascendancy Model (GAAM) identified four key types of AI (Descriptive, Diagnostic, Predictive and Prescriptive), while Eriksson et al. (2020) proposed the extension of GAAM to Creative Analytics, i.e., AI capable of innovation. It is generally agreed that there are six key types of AI, three of them are currently in use: Descriptive, Diagnostic and Predictive. The other three types of AI will take more time to develop: AI is expected to be able to advise of value-creating actions, delegate decision authority to AI with supervision and function fully autonomously, and make decision with no human interaction (Atsmon, 2023).

The classification or taxonomy of AI matters, because each type performs a different function and the analyst guiding its process needs to understand what the strategist is aiming to achieve.

Descriptive AI can be used for performance analysis like dashboards (Atsmon, 2023), however, it could be also useful for data collection and analysis, systematically and effectively identifying patterns and signals that may be missed by humans (Eriksson et al., 2020, Barnea, 2020). It facilitates insights that inform strategy formulation or deliver early warning signs of threats. Diagnostic AI can organise a portfolio into segments, while Predictive AI can provide a systemic view for decision makers (Atsmon, 2023). Predictive analytics is still in development today, and it uses ML to understand human behaviour (Gudigantala et al., 2022). ML is a subset of AI, it depends on the analyst for judging and guiding the process and assessing intermediate results (Sanclemente, 2022), and this potentially makes Predictive AI more difficult and riskier.

OpenAI like ChatGPT available to us today may affect decision making at a strategic level if managers use it to obtain information, filter and organise options, and obtain recommendations in specific situations. Generative AI can assist with data analysis that might guide evaluations, and entrepreneurs and managers may be able to make better judgements as a result of using more data and educated reasoning (Korzynski et al., 2023). However, ChatGPT or other AI is currently incapable of reasoning or decision-making.

With a development of Strong AI, ML would enable the ability to think and operate as ‘human-inspired’ or ‘humanised AI’, only, current AI is still a long way from Strong AI capability (Hesel et al., 2022).

AI and humans: a matter of trust

Considering current or future use of AI for strategic analysis or SDM requires a more in-depth understanding of the interconnection between the human and technology.

Analysts can tell a machine what they want it to do, such as creating decision rules or giving it datasets. Alternatively, analysts can define the output or the goal and let the machine find a solution. In other words, analysts ‘feed’ the task to AI. Framing of the task can influence the SDM process profoundly (Wu et al., 2023). AI mechanisms programmed by humans incorporate human bias in ML: “there is an essential connection between the way the machine performs its actions and how the person behind the machine wants the action to occur” (Sanclemente, 2022, p. 1331). Evidence suggests that such bias may have serious implications, including national security. Mishaps of AI use were reported in the US House of Representatives in 2018, which concluded that considerable flaws and biases can exist in the algorithms that support AI systems. Faulty algorithms and skewed data provide blind spots which raise unmanaged security risks (Sanclemente, 2022).

Another challenge is in sourcing technology talent who can build AI tools and be able to translate business problems into AI questions. For this, analysts need to understand what the company is trying to achieve (Atsmon, 2023; Gudigantala et al., 2022).

From the AI user perspective, a previously mentioned ‘algorithm aversion’ problem is evident. The results of the survey of 500 high-level B2C managers from 2000 of the biggest public companies in the world suggested that human decision makers accept algorithm more frequently when they can modify its decisions or forecasts (Hesel et al., 2022). This implies either a lack of trust in AI solutions or a pre-existing bias. It is not clear yet how grounded the fear of AI is, but the new and emerging AI tools have a high potential to improve ‘decision intelligence’ for better outcomes. The role of humans is unlikely to become redundant with the advancement of AI/ML.

Strong strategic thinking skills combined with sophisticated analytical tools will become a critically important competency of CEOs. It is anticipated that the demand and significance of higher-order thinking skills (i.e., analysis, reasoning, problem solving and decision making) will continue to grow at the global level (Joynes et al., 2019).

AI still needs to be closer to a human ability to solve problems that are not well defined. AI can only perform when it knows what to look for, for example, patterns in a large volume of data to identify insights that can help executives inform their SDM. However, a strategic ‘problem’ or goals need to be articulated first. When a strategy is articulated, then a company needs to decide which type of AI should be used. A real-life example of a Swedish waste management company shows that their sustainability-based strategy was defined first, and then their robotic-based AI system automated their operations which enabled the company to increase their efficiency 20 times and saved them 20,000 euros per month in recycling efficiency (Gudigantala et al., 2022).

Currently, AI is more of a tactical rather than a strategic tool, and AI/ML have a potential to support SDM or be used as building blocks of the strategy to create a competitive edge.

What does the future hold for AI and SDM?

Due to the increasingly complex and ambiguous business environment, the shift from certainty to chaos impacts SDM. In contrast to operational decisions, strategic decisions are pivotal and often irreversible. Traditional decision-making tools such as Monte Carlo simulation, NPV, Decision Trees or Portfolio Optimisation were created for a stable environment, while novel tools like Influence Diagrams, Scenario Planning, Real Options Theory, Systems Thinking or Learning Culture are better suited to organisations experiencing ambiguity and chaos (Wu et al., 2023). The future task of AI will be to enhance the learning capability of ML, so that machines find an optimum solution for us.

There is a consensus of the views that AI will develop predictive analytics which will enable AI to anticipate future events, select best options and scenarios, calculate probabilities like decision trees and in this way inform SDM (Eriksson et al., 2020).

The challenge of formulating the strategic direction falls to senior executives and depends on effective leadership. Because of its complexity, strategy would be one of the later domains to be affected by automation (Atsmon, 2023). Providers of executive education could contribute to developing higher-order cognitive skills and the ‘decision intelligence’ capability of current and emergent leaders.

Conclusion

Strategy formulation is a highly conceptual process, underpinned by strategic thinking and decision making. It requires an understanding of the context and exercising judgment in an increasingly complex VUCA world. Currently, AI is already capable of automating operational decisions, identifying patterns, and providing insights with limited prediction. Still, it needs to be capable of solving undefined problems and generating options for corporate strategies that need to be clearly formulated.

Use of AI raises concerns over risks of over-reliance on AI and probes the trust levels of managers delegating full control to AI/ML. The use of AI/ML is claimed to de-bias decision-making, only ironically, it depends on an analyst setting the parameters for AI tasks, creating a spiral loop of machine-human interaction.

Even with the advancement of strong AI development “humanised AI” will not become a substitute to effective leadership any time soon. Moving forward, the role of executive education will transform from knowledge creation to developing AI-human collaboration capability, enabling executives to think critically and use judgement. Strong strategic thinking skills combined with sophisticated analytical tools will become an essential competency for leaders making strategic decisions for their organisations.

Reference List

Atsmon, Y. (2023). “Artificial intelligence in strategy”, McKinsey & Company, <https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/artificial-intelligence-in-strategy#/>, viewed 22 July 2023.

Adzeh, K.J. (2017). “Strategic Leadership: An Empirical Study of Factors Influencing Leaders’ Strategic Thinking”, American Journal of Business and Management, 6(1), pp.1-15.

Barnea, A. (2020). ‘How will AI change intelligence and decision-making?”, Journal of Intelligence Studies in Business, 10(1), pp. 75-80.

Bennett, N. & Lemoine, J. (2014). “What VUCA Really Means for You”, Harvard Business Review, Jan-Feb, p. 27.

Chadha, S. (2017). “VUCA World: provoking the Future”, Human Capital, January, pp. 14-18.

Duijne, F., van & Bishop, P. (2018). “Introduction to Strategic Foresight”, Future Motions, Dutch Future Society, available online <http://www.futuremotions.nl/wp-content/uploads/2018/01/FutureMotions_introductiondoc_January2018.pdf>, viewed 23 July 2023.

Eriksson, T., Bigi, A., & Bonera, M. (2020). “Think with me, or think for me? On the future role of artificial intelligence in marketing strategy formulation”, The TQM Journal, 32(4), pp. 795-814.

Gudigantala, N., Madhavaram., S., & Bicen, P. (2022). “An AI decision-making framework for business value maximization”, AI Magazine, (44), pp. 67-84).

Hesel, N., Buder, F., & Unfried, M. (2022). “The Next Frontier in Intelligent Augmentation: Human-Machine Collaboration in Strategic Marketing Decision-Making”, NIM Marketing Intelligence Review, 14(2), pp. 49-53.

Hines, A. & Bishop, P.C. (2013). “Framework Foresight: Exploring futures the Houston way”, Futures, 51, pp. 31-49.

Hoglund, L. & Martensson, M. (2019). Entrepreneurship as a Strategic Management Tool for Renewal – The Case of the Swedish Public Employment Service”, Administrative Sciences, 9 (4), 76-86.

Joynes, C., Rossignoli, S., & Fenyiwa Amonoo-Kuofi, E. (2019). “21st Century Skills: Evidence of issues in definition, demand and delivery for development contexts (K4D Helpdesk Report)”, Institute of Development Studies, Brighton, UK.

Korzynski, P., Mazurek, G., Altmann, A., Ejdys, J., Kazlauskaite, R., Paliszkiewiwicz, J., Wach, K., & Zmieba, E. (2023). “Generative artificial intelligence as a new context for management theories: analysis of ChatGTP”, Central European Management Journal, 31(1), pp. 3-13.

Larson, E. (2022). “How decision intelligence will finally change decision making from mystical to mundane”, Forbes, 21 March 2022.

Leavy, B. (2023). “Understanding the fundamental economics of AI”, Emerald Publishing Limited, Strategy and Leadership, 51(2), pp. 17-23.

Liedtka, J.M. (1998). “Strategic Thinking: Can it be Taught?”, Long Range Planning, 31(1), pp. 120-129. 

Martínez-Plumed, F., Gómez, E., & Hernández-Orallo, J. (2021). “Futures of Artificial Intelligence through Technology Readiness Levels”, Telematics and Informatics (58), 101525.

Mintzberg, H. (2000), The Rise and Fall of Strategic Planning, Pearson Education: Great Britain.

Pisapia, J. (2011). “Strategic Leadership: Key Definitions”, <http://johnpisapia.com/think/content/strategic-leadership-key-definitions>, viewed 22 July 2023.

Pisapia, J., Reyes-Guerra D., & Coukos-Semmel, E. (2005). “Developing the Leader’s Strategic Mindset: Establishing the Measures”, Kravis Leadership Institute, Leadership Review (5), pp.41-58.

Ruppanner, L., Churchill, B., Bissel, D., Ghin, P., Hydelund, C., Ainsworth, S., Blackman, A., Borland, J., Cheong, M., Evans, M., Frermann, L., King, T & Vetere, F. (2023). 2023 State of the Future of Work. Work Futures Hallmark Research Initiative, The University of Melbourne.

Sanclemente, G. L. (2022). “Reliability: understanding cognitive human bias in artificial intelligence for national security and intelligence analysis”, Security Journal, (35), pp. 1328-1348.

Wu, C., Ramamohanarao, K., Zhang, R., & Bouvry, P. (2023). “Strategic Decisions: Survey, Taxonomy, and Future Directions from Artificial Intelligence Perspective”, ACM Computing Surveys, 55(12), pp. 1-30.