It is a fair question and it is usually answered badly, either with a shrug or with a number that arrives before anyone has understood the work.
What actually drives cost
- How many systems have to talk to each other, and how well they are documented
- How clean and consistent the existing data is
- How many exceptions the process has, because exceptions are where the hours go
- How much change the people involved have to absorb
Notice that none of those are about AI. The model is rarely the expensive part. The plumbing and the edge cases are.
Judge it against the return, not against a market rate
A workflow that returns ten hours a week to a team is worth a very different number to one that saves an hour a month. Before agreeing a price, agree the number it has to beat.
A cheap build that nobody uses is the most expensive option on the table.
Watch the running cost
Build cost is visible. Running cost is not. Usage, hosting, support and the time someone spends maintaining it all continue after go live, and they belong in the decision from the start.
- Should we start small or do it properly?
- Start small on scope, properly on quality. One workflow built to hold up beats four fragile ones.
- How quickly should it pay back?
- For the work we do, most practical automations should be paying for themselves inside a year, and often much sooner.