Is AI replacing the chief strategy officer?
Research from professors Vikas Mittal and Alessandro Piazza finds that generative AI now performs much of the traditional work of chief strategy officers (CSOs). Their argument: producing a strategy summary is only part of the job. Executives still need evidence of which initiatives improve customer value.
Picture a chief strategy officer who returns from vacation to find that the CEO, in her absence, had an intern run the company’s strategic analysis through Claude. It comes back with industry trends, competitor profiles, market summaries and a set of initiatives with implementation metrics attached. The work is completed in a fraction of the time. It would normally take her team weeks. The scenario Mittal and Piazza pose at the start of their new paper is no longer far-fetched. In a survey of 7,000 professionals, 87% of senior executives reported using AI on the job, compared with 27% of frontline employees. Can a chief strategy officer survive a tool that does the work faster and at comparable quality?
Professors Mittal and Piazza discuss their findings.
What can generative AI already do that used to belong to the chief strategy officer?
AP: The CSO typically wears seven hats: the advisor, the monitor, the dealmaker, the implementer-in-chief, the chief of staff, the special projects leader, the thinker, muse and futurist. If you look closely at the daily work inside those roles, much of it is precisely what large language models (LLMs) are good at: gathering information, summarizing it, synthesizing it and fitting it into an established framework. The advisor and monitor roles are especially vulnerable. The thinker, muse and futurist are the least exposed because they require judgment about futures that current information alone cannot settle. The dealmaker and the chief of staff are harder to automate, because they run on negotiation, relationships and political judgment.
VM: We wanted to know how far this had gone in practice, so we talked with 20 CEOs and strategy officers at 11 companies valued between $10 million and $100 billion. We also sat in on their planning meetings and strategy retreats. Every single company was using gen AI for strategy work. All the executives reported using it to apply frameworks such as SWOT. Ninety-five percent used it for market trend analysis, 90% competitor analysis and 80% financial projections. Ninety percent of the executives told us they expected major parts of the strategy officer’s role, as currently configured, to be replaced. One CEO said their strategy officer used to sit in every high-level meeting, abstracting insights and action priorities, and now an AI tool summarizes the meeting and drafts the priorities.
So is AI replacing the CSO position?
VM: The more uncomfortable question is, what exactly is being replaced? The link between strategy planning and financial performance has always been much weaker than executives assume. An analysis cited in our paper examined 717 published correlations. In that analysis, strategic planning showed a zero or negative correlation with financial outcomes 79% of the time. A separate analysis found 43% of executives doubtful about the effectiveness of their own company’s strategy. One of the CEOs we interviewed said that for the resources his company puts into strategy, the financial return is questionable.
AP: The technology also has limits. Take information synthesis, for example. AI can synthesize quickly, but a lot of what it returns is what we would call “trendslop”: generic summaries that read well and say very little. Automating synthesis does not answer the question a CEO actually has: Which of our strategic initiatives are creating a lift in customer value and which are draining resources? That’s a question of inference, and synthesis is not inference.
What does that leave for the chief strategy officer?
VM: A generative AI summary alone cannot establish which inputs drive a specific outcome, such as customer value. Forty years of studies show a strong association between customer value and outcomes ranging from retention and pricing power to gross margin and firm value. But you have to identify the specific drivers, and that requires company-specific data and statistical modeling, which can be supported by analytical AI. We worked with the strategy officer at a large manufacturer whose team had been scraping competitor websites and buying market sizing reports. The chief commercial officer told us the group produced binder after binder on competitors, and still had no idea what their own customers valued. The model we built with them showed that a one-point increase in customer value was associated with a $936,000 increase in annual sales and a 2.1% increase in gross margin. Of the 87 possible drivers, 12 accounted for more than 90% of the lift in customer value.
AP: And then the executive team has to decide what to do with that. In that case, they chose three drivers to excel on: acknowledging customer problems within an hour, keeping customers informed during the resolution process and resolving problems to the customer’s complete satisfaction. Over the next three quarters, the company reached 91%, 98% and 87% on those goals, respectively, and improved its sales and margins relative to its peers.
VM: In that arrangement, the strategy officer’s authority changes character. It no longer rests on persuading colleagues that she has sound judgment but rather on owning the strategic model built on science, and on being able to say whether an initiative is still earning its place as a priority.
AP: The CSO owns the model, but the executive team jointly owns the conclusions. The CFO interrogates the choice of outcome variables. The COO asks whether the high-lift drivers are operationally manageable. Everyone signs off on trade-offs. In our experience, the longest part of this work is building the capacity of operating executives to act on what the model says, which tends to be a multiyear project rather than a communications exercise.
VM: In short, CEOs already have tools that summarize the market. But they still need someone who can tell them, with evidence, which of their initiatives is actually working.