← All insights

Carousel, 8 slides

A claimed AI saving clears three questions before it is a number.

Technology, operations and the numbers, read as one thing.

AI produces, but producing is not productivity, and without being deliberate about viability, write-offs are more likely than pay-offs.

Even before AI, technology improved outcomes only when the rest of the operating model moved with it. AI just makes that urgent rather than neat.

Three years of AI cases in M&A, across the US, Europe, Asia and the Middle East. A claimed saving has to clear three questions before I would credit it in diligence, or back it in a value creation plan.

Slide 1 of 8. A claimed AI saving clears three questions before it is a number. Technology, operations and the numbers, read as one thing. Slide 2 of 8. Question one: can the business say what right looks like, in advance? Precisely enough that something other than a person can tell when it is wrong. Slide 3 of 8. Question two: does the saving survive crossing a border, and is the model built for that market? Slide 4 of 8. Question three: does it reach the P&L? Consumption pricing, checking in the function that owns the work, maintenance and security, all incremental to the AI invoice. Slide 5 of 8. Auto parts distribution: same business, different markets. What right looks like changes with the channel mix. Slide 6 of 8. Telco: 115 candidates, three went to production. Network energy, fraud and identity, tier-1 support. Slide 7 of 8. Where the work sits, and what still has to clear: a matrix of whether right can be written down in advance against the cost of catching a wrong answer. Slide 8 of 8. The trade: first principles up front, or the hold period gets spent redoing work already paid for. Mechanics over models.

Download the PDF

Testing an AI case in a deal you’re working on? Tell me about it, or read more insights.