The company that pays people to spend on AI

Some companies now set minimum AI spend per employee.

Not maximum. Minimum.

Thousands per person per week on tokens, mandated, to force usage. Meanwhile, the same firms burn seven figures a month on the best model money can rent, and still can't point to the return.

Read the pairing again, because it's a diagnosis. A usage quota is management admitting the tool doesn't pull people toward it. The token bill is management assuming the model was the investment. It wasn't.

Now look at who the quota is aimed at. Knowledge workers, by definition and unlike physical workers, can't be checked at work, only by result. And they deliver the result, today, without AI: the work gets done, the old way works, nothing is failing from their side of the desk. A new instrument, especially one this complex, gets adopted when it visibly pays back the time invested in mastering it, not when it's ordered. Most people, reasonably, won't invest hours into a tool whose benefit they can't see. They'll use it occasionally, when it's handy or interesting, and hit the quota dutifully. Spending happens. Value doesn't.

And notice what they can't do, even if they wanted to: redesign the operating process around the model. That takes a mandate nobody at their desk holds. So the AI sits beside the workflow, waiting to be visited, while the workflow runs exactly as before. The missing investment was never the tokens. It's new architecture and processes rebuilt so the model sits natively inside the work, and people are shown, on their own tasks, where it actually pays.

Buying capability is a purchase order. Absorbing it is a transformation, and it needs an owner with a mandate. The 95% keep discovering the difference.

That owner, and the architecture behind them, is what my book is about.

Has your firm measured AI usage, AI spend, or AI return? Be honest about which of the three.