Take one agent with one job: it reads incoming support tickets, checks the customer's contract and history, and proposes or sends the resolution. The same agent, doing the same work, can arrive on three different bills, and the three bills are not three prices. They pay the supplier to do three different things, and a buyer who picks on the headline number picks the supplier's incentive without reading it.
The first bill is the one the software industry already knows. The agent is priced per user, per month, like a licence. Fifty support staff, fifty seats, whether the agent handled ten tickets or ten thousand.
What it pays the supplier for is the agent existing. Not working, not being right. The supplier's incentive is to sell seats and keep them renewed, which means the incentive is to make the agent look valuable at renewal time, not to make it handle more or handle better. A seat price also has nothing to say about the operations line, the ongoing work of keeping the connections, permissions and trail right, which is the largest cost of running an agent in the first year and which lands on the buyer unbilled.
Per seat is familiar and easy to budget, which is why procurement likes it. It is also the bill least connected to anything the agent does.
The second bill is the one the platform vendors have moved to for support agents. The agent is priced per conversation it resolves without a person. A thousand resolutions a month at a unit price, and the bill rises and falls with volume.
What it pays the supplier for is closed tickets. That is closer to the point, and it has a well-known failure: the supplier is paid when the ticket is marked resolved, not when the customer's problem is. An agent tuned for this bill closes tickets. It answers the question that was asked, marks it resolved, and the customer opens another one, which is another resolution. The number that justifies the bill, resolutions, is also the number the supplier controls.
Per resolution is honest about volume and dishonest about quality, unless the contract defines a resolution as something the customer agreed to and measures reopens. Most do not.
The same shape appears outside support. An agent billed per quote drafted is paid to draft quotes, not to draft the quote that wins. An agent billed per record enriched is paid to touch records.
The third bill is the one the enterprise firms are moving to and the mid-market is starting to be offered. The agent is priced against a business result: first-contact resolution rate above a threshold, backlog below a number, average handling cost down by a percentage. The supplier is paid when the result holds, and less or nothing when it does not.
What it pays the supplier for is the thing the buyer actually wanted. That is the right principle, and it comes with two conditions that decide whether the principle is real.
The first is the baseline. An outcome has to be measured before the agent starts, by a method both sides accept, or the outcome is whatever the supplier's dashboard says it is. A supplier who offers outcome pricing without insisting on the baseline is offering per seat with a bonus.
The second is the operations underneath. Keeping a result true month after month is operations work, and someone is paying for it. If the outcome price does not include it, the supplier will keep the result by asking the buyer for the work, which is the hours model wearing a different label. We set out how the three pricing models for human consulting pay for three different things in how Zoho consulting services are priced; the three agent bills are the same story with a model in the loop.
Run the support agent for a year on each.
Per seat: a fixed monthly number, unchanged whether the agent was useful, with the operations cost arriving separately and unplanned. Easy to budget, impossible to judge.
Per resolution: a number that tracks volume, which is good, and that the supplier can grow by closing faster, which is not. Judgeable only if reopens are measured.
Per outcome: a number tied to the result, judgeable by construction, and only honest if the baseline exists and the operations are inside the price.
The buyer's question is not which is cheapest. It is which one pays the supplier to do the thing I want, and what it costs me to check.
The failures are specific enough to name. Under per seat, the common outcome is an agent switched on for fifty people and used by six, with the renewal justified by a slide about adoption. Under per resolution, the common outcome is a rising resolution count beside a rising reopen count, and a customer satisfaction score that nobody put in the contract. Under per outcome without a baseline, the common outcome is a dispute in month four about what the number was in month zero, settled in the supplier's favour because the supplier has the dashboard.
None of those is a bad supplier. They are the bill working as designed. The design is what the buyer is choosing.
We price agents as a managed subscription with a usage tier and an outcome condition, which is a hybrid, and it is a hybrid on purpose.
The usage tier is the honest version of per resolution: the price moves with the volume the agent handles, because tokens and operations move with it, and the tier includes the operations line rather than leaving it to arrive later. The outcome condition is the honest version of per outcome: a number measured before we start, read every month, with one party accountable for it, and with the agent's boundary widening only as the trail shows the number holding. Guaranteed estimates cover the setup, so the first bill is the bill.
We chose it because it is the only shape where our incentive and the buyer's point in the same direction for the whole life of the agent, not just at renewal, and because we have seen the other two fail in exactly the ways described above. It is the same reason a company buys accountability from a fractional CTO rather than hours, which we wrote up in what a fractional CTO costs and what the fee buys.
What does this bill pay you to do? What is the baseline, who measured it, and when? Is the operations work inside the price or outside it? What happens to the bill when the agent is wrong?
A supplier with good answers to those four is selling an agent. A supplier without them is selling a bill.
Discovery is no-risk: we take the baseline for the job your agent would do, show you the three bills against it, and you pay only if you go ahead.
In a no-risk discovery we look at your CRM, finance and operations systems and who owns each part, and show what a connected system would do differently. You pay only if you proceed. Or see how we approach it.
More on the same problem:
"Something isn't working." "We're outgrowing our tools." "We're afraid to make the wrong move." No-Risk Discovery is a short, practical conversation that gets you clarity before you commit to anything big. We'll tell you if we're a fit. If we're not, we'll tell you that too.
Say who you are and ask. A person replies within one business day.
