Every AI transformation that fails inside an established company fails for one of these three reasons. Hold the line on all three before you spend a euro on Layer 1.
Not a one-off speech at the kick-off. Visible, repeated, public backing — every quarter, in every all-hands, on every spend review.
The minute the CEO and exec team go quiet on AI, the middle of the org defaults to the old way. Layer 2 dies in the budget cycle, Layer 3 never starts. Without ongoing top-down conviction, even the best champions burn out.
Litmus test: name the next three exec moments where AI will be on the agenda.One named person, full mandate, exec air-cover. Not a committee. Not a project office. Not the IT director by default.
The champion holds the roadmap, picks the pilots, kills the dead ones, and answers to the CEO directly. They have permission to say no to good ideas that aren't the priority — and the political standing to make those nos stick.
Litmus test: when something stalls, who has both the authority and the calendar to unstick it?Ambition for the destination. Discipline at the start line. One process, one team, one measurable win — before week twelve.
The vision is a 35-year-old chemicals trader becoming AI-native. The first move is one trader using Claude on Monday. Don't launch ten pilots; launch one and let it succeed loudly. The next nine fund themselves once the first one ships.
Litmus test: what is the single specific win you'll point to in 90 days?All three. Two out of three is not enough.