BCG's Line for AI-First COOs: The Constraint Isn't the Technology, It's Whether You Know What You Want From It

“The constraint will not be what technology can do, but whether COOs know what they want it to do.” That line, from BCG’s latest entry in its “AI-First Senior Leadership Team” series, reframes the COO hiring question CEOs are increasingly facing: not whether a candidate understands AI tools, but whether they can redesign operations from first principles instead of automating the workflow that already exists. BCG estimates comprehensive AI transformation could lift industrial operations productivity more than 30% within two to three years, and potentially triple it within five — a spread wide enough that the difference between those outcomes is almost entirely the operating model choice, not the technology available.

BCG names five capabilities an AI-first COO needs beyond traditional operational strengths like pattern recognition and operational discipline: a different decision-making framework, a bias toward reinvention over optimization, a shift from reactive to predictive operations, AI fluency that doesn’t override human judgment, and the ability to build technical capability across a team that didn’t previously need it. The piece gives CEOs three assessment questions to tell a genuinely transforming COO from one who’s layering AI tools onto an unchanged process.

That distinction — stalled by unclear intent rather than by the technology — echoes BCG’s own earlier framing, which named “AI paralysis” as a specific failure mode distinct from simply moving slowly: teams stop advancing not because leadership won’t support the effort, but because they lack the clarity to act on it. Read together, BCG’s throughline across roles is consistent: the bottleneck in enterprise AI transformation has moved from capability to intent, and it shows up first in whoever owns operations.

For engagements evaluating operations leadership or AI operating-model maturity, that’s a sharper diagnostic than “does this person know AI” — it’s “can this person say, specifically, what they want it to change.”