Open your profit and loss statement. Every operating cost on it is, one way or another, a bundle of human hours: the CSR answering phones, the dispatcher juggling the board, the marketer writing the campaign, the bookkeeper closing the month. For three hundred years, the unit of work has been the human hour, and every business model we know — including yours — is built on top of that unit.
That unit is being replaced. Not everywhere, not overnight, and not for the work that happens with two hands inside a condenser. But for a widening band of the work that surrounds the truck — answering, scheduling, following up, writing, reporting, analyzing — the cost of execution is shifting from hours to tokens: metered units of machine intelligence you buy the way you buy fuel.
What changes when the unit of work changes
A unit change is not a discount — it is a different physics. Three properties of token-priced work make it behave unlike anything on your current P&L:
- It scales instantly. Hiring an hour of human work takes weeks of recruiting. Buying another million tokens takes a keystroke. Capacity stops being a constraint you plan around and becomes a dial you turn.
- It never sleeps. The after-hours call, the 2 a.m. web visitor, the Sunday estimate request — work priced in tokens is awake for all of it. Availability stops being a staffing decision.
- It gets cheaper while getting better. Human hours get more expensive every year — rightly. Compute has moved the opposite direction for decades. You are watching the only input in your business with a falling cost curve and a rising capability curve.
The mistake: treating this as a cost story
The lazy reading is "AI makes things cheaper." That reading gets owners in trouble, because the first company in a market to convert hours to tokens does not pocket the difference — it spends the difference on speed, availability, and follow-up, and starts winning jobs the slower companies never even saw. I have said versions of this from stages for years: the companies that treat AI as a cost-cutting tool will be outrun by the companies that treat it as a capacity multiplier. On the Kraft Your Life Radio Show I put it plainly: the operators who move first won't just be more efficient — they'll be reachable, responsive, and present in ways the hour-priced competitor cannot afford to match.
Questions to take to your next leadership meeting
- Which of our workflows are pure information work — no wrench required — and what do they cost us today in loaded hours?
- If a competitor ran those same workflows on tokens at a fraction of the cost, what would they do with the savings: cut price, or out-market us?
- What is our cost per answered call, per booked job, per produced piece of content — and do we even measure those in a way that would show the shift?
- Who in our company owns the token budget? (If the answer is "nobody," that is the answer.)
Where this series goes
This essay opens the From Labor Hours to Tokens pillar — the economics track of this publication. Coming installments get concrete: capacity planning when one operator runs a fleet of agents, pricing when your delivery costs collapse, and the ratios your CFO should be watching. If you want the philosophical version — what this shift means for the people doing the work, not just the spreadsheet — that story is in Hands Up, particularly chapter ten, "Optimize for Both."
The hour built your company. It will not build the next one. And if you want the lived proof instead of the theory, the case study is my own P&L: from 320 people to 38, over $10 million in revenue, on an AI intelligence layer.