How to Build and Pressure-Test Your Strategic Vision - Where ChatGPT Can Help
…and where it cannot
[Source: Will AI replace management consultants?, Urvesh Karia]
The media spotlight on generative AI tools passing Wharton MBA exams has revived the concern of whether machines can replace managers. Business strategy though seems to be more immune than other disciplines (for now). In a recent survey by McKinsey & Co, 25-30% of respondents confirmed the use of AI in areas like marketing, supply chain, and service operations, but only 7% said they used it in strategy or even financial planning. This seems reasonable because the process of articulating how to most effectively combine a firm’s resources and activities with external conditions to capture value, create a sustainable competitive advantage and drive growth is highly integrative, conceptual, and inherently messy.
A hypothetical company called, say, LillyLemon can serve as an illustration. LillyLemon is a longstanding player in the women's fashion market, battling declining revenues each year for the past five years. Pressure testing and refining its strategic vision to drive more revenue, a seemingly amorphous exercise, would however lean on certain timeless principles of strategy:
Frame & diagnose
In their book, Making Great Strategy, Stanford professors Jesper Sørensen and Glenn Carroll advise starting with the destination (or conclusion) when mapping your strategy argument. In this case, it would then make sense to start with LillyLemon’s end goal of driving more revenue and working backwards to identify the set of conditions that would enable the achievement of that goal. Without this first building block of disciplined deductive reasoning in place, it would be a challenge for the CEO to understand what the right strategies for the company should be at their current inflection point.
Forecast and commit
Armed with insight on LillyLemon’s starting position, its CEO would also need a perspective on how the future may unfold. This would help the company develop and explore potential options for increasing revenue to eventually decide on the alternative which would maximize movement towards their end goal.
In order to shape these insights into robust strategies and cement buy-in, the strategists at LillyLemon would not only require rigorous analysis of the ROI and tradeoffs associated with each alternative but also nuanced interpersonal engagement and debate from senior executives. Carroll also advocates that the debate process includes a “suspension of disbelief” to determine which of the proposed alternatives would be valid today and prove to be the sound choice tomorrow.
Monitor & pivot (as needed)
The final imperative is to constantly monitor and refresh the strategy as conditions change and new information becomes available. Good strategists track the progression of a strategic recommendation from its roots as an idea through its emergence as an operational reality, and need the ability to reassess a charted path often in the midst of ambiguous, highly charged situations. Cutting through the information overload and articulating the known unknowns is a critical part of reacting in an agile manner.
Could ChatGPT (or generative AI more broadly) replace strategic insight?
In our hypothetical example, could ChatGPT decide, in place of LillyLemon’s CEO and her strategy consultants, what the right strategy is? Probably not - it’s hard for AI to know everything an executive knows. But there are opportunities to use AI in the building blocks of strategy that could significantly augment outcomes.
1. ChatGPT could look backwards at the business to understand the key drivers of performance, with a velocity that would accelerate analyses that are key inputs into strategy.
In terms of diagnostic intelligence, ChatGPT can organize a business’ portfolio into segments to understand iteratively at a granular level where performance is coming from much more rapidly than a team of analysts could. This, in turn, would allow the analysts to scan for certain patterns and create a significant strategic edge for them through a relatively tactical tool.
In LillyLemon’s case, analyzing the root causes of declining revenues would entail understanding the key revenue drivers for the company, assessing the market it operates in and consumer preferences in that market. Gaining insight into these conditions would further track back to a more granular set of conditions as to whether it would serve LillyLemon better to raise prices or to increase volumes.
2. ChatGPT could also provide a systematic viewpoint to look forward - and to revisit big decisions that were based on assumptions about the world that may have since changed.
Good strategy influences optimal resource allocation. The predicted ROI of strategic initiatives influences how a company deploys executive time and talent and how it spends money and focuses sales efforts. For example, shifting styles to increase customer acquisition might be a path forward for Lily Lemon if an analysis of consumer preferences were to indicate evolving clothing sensibilities among a large proportion of the target market. If the data were to suggest that demand was inelastic implying a higher willingness to pay for quality clothing (in light of the use-and-throw fast fashion culture of competitors), raising prices could be a viable alternative. Or the data could point towards a third alternative which would change neither merchandising nor pricing strategy, but would instead increase volume through sales and advertising.
Generative AI can provide an objective prediction of performance starting from a neutral momentum case - i.e., based on everything that happened in the past and some indicators about the future, what is the forecast of performance if we do nothing? This kind of predictive intelligence can change the dynamics of the resource allocation discussion and provide a baseline for better decision-making. Like most of the value we draw from AI, strategists shouldn’t treat this as definitive. But tracking the evolution of the assumptions behind a prediction and understanding why a specific prediction was made allows them to more effectively assess whether to trust the prediction or not.
3. By being dispassionate and purely analytical, ChatGPT can counter human biases in strategy development that are “systemic, observable and predictable”.
Most strategists have, at some point, been in a strategy development session where the CEO proposes something and everyone accedes without debate or discussion, or every business unit leader advocates for their group to get the most resources, or the same executives make different decisions on December 31 and March 1 to make the numbers. AI could encourage devil’s advocacy by highlighting a fact-based reality and the human biases that inevitably creep into decision making.
In our LillyLemon scenario, the CEO might have initially approved a new sales and advertising strategy deploying a rewards program giving customers a 20% discount on all purchases after paying a one-time activation fee to join the program. However, testing the in-market rewards program might suggest the need to refine the proposed strategy.
A purely objective analysis run by AI immune to hierarchy or to the cherry-picking of data to corroborate pre-existing ideas or decisions can allow viewpoints contrary to the management hypothesis, enabling a richer and more balanced debate. The CEO still decides; AI can just provide that crucial non-subjective and impartial filter.
***
“When you are looking at the latest tool, you need to consider how it can integrate into your business, and what it can do for you that furthers or complements what you are already doing.”
As the Wharton School marketing professor, Raghu Iyengar points out, AI capabilities are unlikely to decide, in place of the business leader, what the right strategy is. For one, AI systems require huge amounts of labeled data on past performance to perform analyses accurately, which may simply not be available in some domains. Secondly, AI cannot create an output or provide a recommendation outside of what is a recognizable pattern. LillyLemon’s strategists, for example, would still need to determine how to approach their customers in novel ways after distilling the root causes of lackluster revenues (with help from AI). Lastly, lateral thinking across domains would still lean heavily on strategists cross-pollinating; to decide for example how the solution for revenue growth for an apparel retailer could translate into potential learnings for another consumer products company in another geography using distribution channels different than LillyLemon’s.
Deploying generative AI more extensively is likely to liberate strategists from routine number crunching towards thinking more deeply about what really makes us human. Competitive advantage will increasingly rest in having executives who know how to apply AI in the building blocks of strategy to utilize its strengths while also understanding its limitations.
(Hustle Fuel represents my own personal views. I am speaking for myself and not on behalf of my employer, Microsoft Corporation.)