There is an awkward conversation happening in a lot of vendor renewals right now. A company is paying an outsourcing or services partner by the head, or by the hour, the way it always has. Then it learns the partner has quietly introduced AI into the workflow and is getting through the same volume with fewer people in less time. The work is still billed the old way. The efficiency showed up, but it landed in the vendor's margin, not the client's invoice.
The buyer's question is reasonable and a little uncomfortable. If a machine is now doing a meaningful share of this, what exactly am I paying for? It is the right question, and most pricing models were not built to answer it.
The instinct, once a finance leader notices the gap, is to claw the savings back by cutting the headcount they are paying for. Fewer seats, lower bill. The problem is that this assumes the seats were the thing of value, when often the seats were just the unit of measurement. You can cut your way to a smaller invoice and a worse outcome at the same time, which is how companies end up rehiring six months later.
The more useful move is to change what you are buying, not just how much of it. For most of the history of outsourcing, you bought labor: a number of people, or a number of hours, and you hoped the work that came out the other end was good. That made sense when the cost of the work scaled almost perfectly with the number of humans doing it. AI breaks that link. When a tool can absorb a large share of the volume, paying per person stops describing the value you are actually receiving.
A finance leader described almost exactly this to me last year, and she was not the first. The details below are blended; the dynamic is real.
A mid-market company runs a high-volume back-office function through an outsourcing partner, priced per full-time person. Over time, the partner gets faster. Tickets that used to take a team a full day are clearing in hours. The client's finance lead, sharp and doing her job, sees the throughput climb and the headcount stay flat and concludes she is overpaying. At renewal, she pushes hard to cut the number of seats.
She gets the lower number she asked for. Then quality slips. With fewer people on the account, the unusual cases that always needed a human start slipping through, and the cost of those misses, in rework and in customer trust, quickly swamps the savings on the contract. She had optimized the one number she could see, the seat count, without a way to see or pay for the thing she actually cared about, which was the work coming out correct and on time.
The renegotiation that fixed it did not haggle over seats at all. It redefined the deal around the outcome: a price tied to volume of work completed to an agreed quality standard, with the partner free to hit that standard however it could, including with automation. Suddenly the incentives lined up. If the partner could use AI to deliver the same quality cheaper, both sides benefited, and the client stopped paying for inputs it could not evaluate and started paying for results it could.
The reason outcome-based pricing is having a moment is not that it is trendy. It is that it survives contact with automation in a way that seat-based and hourly pricing do not.
When you pay per hour, you are paying for effort, and you are quietly rooting against efficiency, because every hour the vendor saves is revenue they lose. The incentives are backwards. Neither side has a clean reason to automate, because automation shrinks the very thing being billed. When you pay for an outcome, a resolved case, a processed transaction, or a clean financial close, the vendor is rewarded for getting there more efficiently, and you no longer care whether the work was done by ten people or three people and an agent. You care that it was done correctly, on time, and at a price you agreed to. That is a far more honest transaction, or what a buyer wanted all along.
This is also why the savings question and the quality question have to be answered together. An outcome price only protects you if the outcome is defined to include quality, which means a standard you can measure and a consequence when it is missed. Price the outcome as raw volume with no quality bar, and you have just recreated the Klarna mistake with a different invoice. Tie the price to volume delivered at a defined standard, and you have something that holds up. The enterprise AI results that hold up tend to come from redesigning the work around what the technology can now do rather than bolting a tool onto an arrangement built for the old way.
A managed operation is structurally easier to put on an outcome-based deal than a staffing arrangement, because the partner controls the whole workflow and can be held to a result rather than a roster. It is also where the AI question gets least adversarial. If the contract pays for outcomes, you do not have to police whether or how your partner uses automation or fight over who keeps the savings. You set the standard and the price, and a partner that can pair agents with expert operators to hit it more efficiently is doing exactly what you are paying them to do. The automation becomes their problem to deploy well and your benefit to receive, instead of a hidden variable in a per-seat invoice.
If you are renewing a services contract this year and you suspect AI is quietly changing the economics under you, counting seats harder will not get you there. The better question is what outcome you are actually buying and whether your contract pays for that or for a proxy.
Define the result you want in terms you can measure. Attach the price to that result, with a real quality standard baked in, not just volume. Stop paying for effort you cannot evaluate, and stop trying to capture savings by cutting the inputs that were holding quality together. Done well, you get the upside of automation without having to become an expert in your vendor's tooling, and your partner gets rewarded for the thing you wanted in the first place. The seat count was never the point. The work coming out right was.