The cost of selling the work

I'm building a service that uses AI, and I need to be careful not to price it as though my own time is free.

I like Sequoia's argument about selling the work instead of another tool. The customer gets what they needed without having to do it themselves, and better models can help you get it done more efficiently.

Say I'm putting together a research brief. Getting a draft back quickly doesn't mean I'm finished. If a source is missing, or two sources disagree, I'm still there working out what to trust and what needs changing. Then I need to check whether it actually answers the customer's question.

It's tempting to see all of that as stuff to automate away. Some of it might be. But if the customer wanted to check sources and sort out contradictions themselves, they might have been happy with a tool. Having someone do that work could be why they hired me.

So I'd want to track the whole job, from the request through to the customer accepting it, including my time. And I'd keep correcting my own mistake separate from being asked to answer a different question. I owe them the first. The second needs a conversation about what we agreed to do.

If you're pricing a service like this, include the work you expect to do after the AI gives you something back. That may be a perfectly good service to sell. You just need to get paid for that part too.