In the Room: How CEOs Balance AI Urgency and Risk
AI transformation is 70% a people problem, not a technology one, and CEOs should start with two or three high-impact use cases while explicitly redefining what counts as success.
Summary
BCG senior advisor Tiger Tyagarajan, drawing on decades of CEO-coaching experience, argues that unlike earlier technology waves such as cloud (where companies could safely wait), AI adoption punishes delay with real competitive disadvantage — but reckless speed is equally dangerous. His "10-20-70" framework attributes only 10% of transformation difficulty to technology and 20% to data, with 70% residing in people and change management. He advises starting with two to three high-impact problems rather than enterprise-wide rollouts, and resetting success metrics to treat a 70% success rate as a win given AI's experimental nature.
Editor's opinion
The 10-20-70 framing and the human-error-rate asymmetry are the only two ideas here with any staying power, and even those are asserted rather than demonstrated -- there is no data, case study, or named client example behind any of the three steps, just first-person coaching experience presented as a generic advice column. The “aperture expansion” examples (drug screening, claims processing) are illustrative rather than evidenced, and the closing message (“you’re not supposed to get it exactly right”) is more motivational than actionable. Readable, but too thin to treat as more than a talking point.
Key concepts
10-20-70 rule · aperture expansion