When a Client's AI Usage Doubles Mid-Month

Usage-based pricing is a model in which part of what a customer pays depends on how much of a metered resource they use, and it sits uneasily next to a flat retainer. A retainer promises a fixed monthly number. AI work has a cost that moves with use. Sooner or later a client uses twice what you expected, and what happens next should have been decided before they signed.

Why it happens

It is rarely misuse. A client finds the thing useful and does more of it. A new starter runs the interviews that had been put off. A project starts and ten processes need reviewing in a week instead of one a month.

The same shape appears in your own use. In abi., AI work is measured in credits, and an interview costs 40, a document read 15, a scored call 25 and a review in Agent Studio 25. A plan carries a monthly allowance and the allowance resets each month. Because the costs are published, you can work out in advance what a busy month looks like, and a busy month is not an emergency. It is a Tuesday.

Decide the ceiling first

Before the retainer starts, agree what the included usage is and say it in terms the client can picture. Not "a thousand credits", which means nothing to them. "Up to this many process reviews and this many interviews a month" does.

Then decide, in writing, what happens at the ceiling. There are only three sensible answers.

It pauses. The work stops until the next month, and the client knows that in advance. This is the cleanest, and it suits a client who would rather not have a surprise.

It tops up. The client buys more, at a price agreed in the contract, and the work carries on. This suits a client who wants it done now.

It is re-quoted. A sustained rise in use means the retainer was priced for a smaller job than the one that exists. That is a conversation, and a good one, because the client is telling you in the clearest way possible that the service is worth more to them than they paid for.

Pick the answer that fits each client and put it on the page. What you must not do is decide it at the point the ceiling is hit, because by then you are negotiating with a client who is mid-project.

See it before it arrives

The point of a ceiling is that nobody is surprised. That needs a number you can see. In abi., usage is visible at any time, and the platform stops when the allowance runs out rather than billing for more. That stopping is worth knowing about, because it means the worst case on the abi. side is a pause. It is not a bill.

Build a check into your own routine. Look at usage at the middle of the month, not the end. A client at seventy per cent by the fifteenth is a call to make now, not a line to explain on the invoice.

What credits do not cover

Be careful about what the number measures. Credits measure the AI work done inside abi.: interviews, document reads, scored calls, reviews. They do not measure what a workflow does once it is running. A workflow built in Agent Studio runs in the client's own n8n, and whatever it calls from there is billed where it runs, by whoever runs it.

So there are two running costs in an AI retainer, and a flat price needs to account for both. The first is the work you do in abi. to design and maintain. The second is whatever the live automation uses day to day. If the client already has accounts for the second, say whose they are. If they do not, say that too.

Say it to the client in one paragraph

A client does not want a pricing model. They want to know what happens if they use it a lot. Give them a paragraph: what is included each month, how they will see where they are, and what happens at the limit. If you cannot write that paragraph, the retainer is not ready to quote.

When a flat price is right

Some clients should not be on a metered model at all. If the usage is steady, small and predictable, a flat price with a generous margin is simpler for everybody, and the effort of measuring costs more than the risk it covers. The metered thinking earns its place where use is lumpy, where it grows, or where one big month could wipe out a quarter's margin.

Doubling in a month is an event worth planning for. If you have decided what it means before it happens, it is a pleasant surprise. If you have not, it is the start of an awkward email.

One last habit is worth having. When usage does double, ask what changed before you decide what to charge. Sometimes it is a one-off, such as a backlog finally being cleared, and nothing needs to move. Sometimes it is the first sign that the client has found a second use for the service, and that is the more interesting finding. Either way, the answer shapes the conversation, and you only get it by asking.