BCG: 94% of Chief Strategy Officers Call Innovation Critical. Only 12% Feel Prepared.
Ninety-four percent of chief strategy officers call innovation critical to their company’s future. Only 12% feel prepared for it. That gap sits at the center of a new BCG survey of 175 CSOs examining how the strategist’s job changes once AI takes over the routine analytical work that used to be the job.
BCG frames the shift as four challenges, each paired with a mode of thinking the AI era demands rather than removes. Rising complexity — contextual factors like politics, geopolitics, technology, and climate now account for almost half of the variation in company profitability — calls for systems thinking. Continuous reinvention, illustrated by the Kodak–Fujifilm split, calls for counterfactual thinking. Conflicting time horizons — only 2% of companies successfully balance growth and profitability across a full economic cycle — call for integrative thinking. And AI itself calls for critical thinking: 60% of CSOs already call AI central to strategy, but 75% say they need real capability-building before they can use it well.
The thesis is blunt for a strategy firm to publish: analytical skill has defined the function for decades, and AI now makes analysis faster and more widely available than any one strategist’s judgment. The edge that’s left is knowing which mode of thinking a given problem actually needs.
BCG made a related, sharper argument three weeks earlier about who owns that judgment gap, in its board-relations research. CEOs and boards each believe they’re aligned on AI strategy while actually talking past each other — 61% of CEOs think their boards are rushing transformations, while 75% of board members rate their own AI understanding as equal to or better than their peers. BCG’s Kristy Ellmer put the CEO’s job plainly: “Where are we just keeping up, and where is AI actually creating competitive advantage?”
Read together, the two pieces describe the same capability gap at two altitudes — the CSO’s desk and the boardroom table — and neither closes on its own just because the model gets better.