When Time and Materials Go to Zero

What everyone will be selling once the model is a utility, and why the people who were right about pricing have been right the whole time.

When time and materials go to zero, what will you actually be selling?

I have been asking some version of that question for years. YEARS I tell you. Usually the room nods, agrees it is a great question, and goes back to billing by the hour. Thats because the threat of this question has huge implications. Especially on our friends in the procurement department.

This week the question stopped being rhetorical and started sending invoices.

Two stories ran on the same days and almost nobody filed them together. A Chinese open model, GLM 5.2, landed within a point of the best American frontier model on the benchmarks that actually matter for real work, at roughly a fifth of the price. And Oracle had its worst week in the market since the 2001 dot-com bust, the stock down nearly a fifth, on fears about the debt it has piled up to build AI. The first story got read as "China caught up." The second got read as "the bubble is wobbling." They are the same story, told from opposite ends of the same pipe.

The story is this: the material went to zero.

Both Inputs are Collapsing at Once

Every business that bills by time and materials is selling two inputs. The hours, and the output those hours produce. AI is driving the price of both toward the floor at the same moment. The materials, the raw generated work, a competent model now produces for cents. And the time, the hours of a smart person turning a brief into a deliverable, collapses right next to it. The strategy deck that took a team of mine a month a few years ago can now be generated in about an afternoon. I have written about that part before. The two things that have funded agencies, consultancies, law firms, and most of corporate knowledge work for a century are losing their price at once.

Oracle is the supplier side of the same collapse, and it is worth looking at because the market just flinched at it in real time. Building the capability requires permanent things: data centers, power, a decade of depreciation, billions in debt that has to be serviced on a long, patient schedule. But the thing all of that produces, the capability itself, is a fifth of the price within a year and dropping. You are financing a ten-year asset to manufacture a ten-month one. It is the whole tension of the moment, sitting on one company's income statement, and the investors looking at it got nervous for the correct reason even if they could not name it.

Here is the thing to sit with before the optimism kicks in. T&M was never really a pricing philosophy.

It was a convenient proxy. Everyone agreed to pretend the hours were the value, because the hours were easy to count and expensive to buy. Make them cheap and the proxy stops proxying anything at all.

Efficiency is a Countdown Timer

Which is exactly why the loudest pitch this week, AI as efficiency, is the tell. Do more with less. Expand the margin. Ship it faster. It sounds like progress, and it is, for about one quarter. If you price on inputs and your inputs go to zero, efficiency is just you automating your own invoice downward. You deliver the same outcome with a tenth of the billable hours, and on a time-and-materials model that is not a margin story. Every productivity gain you celebrate is a line item you are training the client to stop paying for. The shops posting about their shiny new AI workflows are livestreaming the demolition of their own business model and calling it a case study.

The Last Pricing Model Standing

Cost-plus pricing has a math problem it cannot survive. The model is your cost times a markup. Run the cost to zero and it does not matter what the markup is, because anything times nothing is nothing. There is exactly one pricing model left that still computes when the inputs are free, and it is the one a small, stubborn band of people [myself included] have been arguing for since long before any of this.

Price the outcome, not the inputs.

And no I dont mean performance marketing. Thats a different story. What I am talking about is charging for what the result is worth to the buyer, not what it cost you to produce. The value-pricing crowd, the people who spent two decades trying to kill the billable hour, were not selling a fancier invoice. They were right about the underlying thing, which is that clients were always buying an outcome and you were always handing them a timesheet as a stand-in for it. AI did not invent that argument. It just ended it. The proxy is worthless now, and the only coherent thing left to price on is the value, and that was the point all along. Or should have been.

I think value-based pricing ultimately wins. It may look a little different or be a different flavor, or answer to a new name, but the spirit will win. Not because it is noble, but because it is the last model standing when the cost floor falls out.

Here is the part I find genuinely elegant. To price on value, you have to be able to do three things a machine producing infinite cheap output cannot do for you.

1. You have to know what the outcome is actually worth to this particular buyer, which is judgment.
2. You have to pick the right output from the thousand competent ones the model just dumped on your desk, which is taste, taste as selection now rather than taste as production.
3. And you have to put your name on the result and stand behind it when it goes sideways, which is accountability, someone to call at two in the morning.

Those happen to be the exact capabilities that do not commoditize, at least not yet. So value-based pricing is not a billing trick you bolt on the side. It is a forcing function. You cannot charge for value you cannot identify, choose, and stand behind, so pricing on value drags you into building the only things that survive the current state of nearly everything. The list of what is left to sell and the list of what value pricing makes you get good at turn out to be the same list.

Where This Could be Wrong

I will give the other side its due, because this is my opinion and I would rather argue it honestly than pretend it is settled. Value pricing is hard. It always has been. Clients resist it, it demands a conversation about worth that most account teams are not built to have, and measurement gets genuinely messy when the outcome is slow or diffuse. And some work really is commodity throughput, where the buyer wants a competent thing cheap and fast and does not care whose name is on it. That work will get priced like the commodity it is, and it should. None of that changes the direction. It changes the timeline. When the cost floor drops out from under cost-plus, the firms that already know how to price on value have a model that works on Monday morning, and the ones that do not have a model that is bleeding out quietly while they congratulate themselves on their efficiency gains.

The bet

So here is the bet. Everyone is about to have the best model. Capability is a utility now, metered and cheap and improving on somebody else's roadmap, and your access to it is a rounding error your competitor can expense by Tuesday. The firms that make it through are not the ones with the best AI. They are the ones who already knew they were selling outcomes instead of hours, who built the judgment and the taste of their people. alongside the willingness to be accountable. When time and materials go to zero, the only thing left to sell is knowing what is worth making, choosing the version that lands, and signing your name to it.