Paying per result instead of per seat
Business software has been sold by the seat for twenty years: a monthly price per user, whatever they do with it. That made sense when software was a tool people used. It makes less sense when software does the work itself.
Why per-seat pricing breaks
If an automated process handles most of the invoices, the number of people logged in tells you nothing about the value delivered. A company could need fewer seats and get more done. Vendors know this, and some have started pricing by usage instead: per request, per transaction, per completed task.
What it means when you buy
Paying for results is often better for the buyer: you pay when work is done, not for access that sits idle. But it moves the risk. A process that runs more than expected, or a system that repeats work because it failed the first time, now costs money each time. Read usage-based contracts carefully: what counts as a unit, what happens when something fails and is retried, and whether there is a cap.
Know your own cost per task
The same logic applies inside your company. If AI does part of the work, its cost belongs with the cost of delivering the service, not hidden in the IT budget. Ask your finance team a simple question: what does it cost us, in AI usage and in people's time, to complete one order, one claim, one support request?
If nobody can answer, you cannot tell whether AI is saving money or just spending it, and you cannot price your own services properly.
What it means when you sell
If your customers increasingly buy by outcome, consider selling that way too. Pricing by result is harder to set up, but it ties what you earn to what the customer gets, and that is an easy argument to make.