BCG's Jeff Walters: Western CEOs Are Watching the Wrong Region on AI Strategy

Western CEOs taking a wait-and-see approach to AI are watching the wrong region, according to BCG Managing Director and Senior Partner Jeff Walters, writing from Singapore. His argument, drawn from Asia-Pacific companies, starts with ownership: the CEO has to personally drive AI strategy rather than delegate it to a technology function, picking the organization’s most important processes and setting explicit, ambitious KPIs — Walters’s benchmark is 50% or more, such as running R&D 50% faster or lifting advertising returns 50%. Delegation, in his read, is where most Western AI strategies quietly stall.

The other four lessons build on that point. Companies need a proprietary “business-context muscle” — consumer insight and process knowledge a competitor can’t copy — and should start building it now rather than wait for better models. Leading firms run more than one foundation model, pairing cheaper “good-enough” models with premium ones by task. Walters names two traps: the “pilot trap” of running thousands of small initiatives that add up to no real result, and the “data trap” of waiting for perfect data instead of prioritizing the critical datasets already on hand, including unstructured sources like meeting notes. He also urges Western leaders to study digital consumer experiences in China and across Asia-Pacific to reset their sense of what’s already normal there.

That instinct — differentiation comes from what a company builds around the model, not access to the model itself — is the same argument Forrester’s Chief Research Officer Sharyn Leaver and Chief Product Officer Carrie Johnson made three days earlier: with AI tooling now commoditized, the real edge comes from pairing it with human judgment and proprietary context, what they call “AI + HI.” Forrester’s own generative-research tool, introduced in 2023, runs on that premise — fast answers grounded in the firm’s proprietary research and analyst expertise, not the model alone. Walters’s “business-context muscle” and Forrester’s “AI + HI” describe the same bet from two vantage points: once everyone has access to comparable models, the compounding advantage is the proprietary knowledge and judgment layered on top — and a company waiting for a better model instead of building that layer now is optimizing for the wrong constraint.