I left €10k+ on the table on my first AI build. Here’s the math I should have done.

Quick lesson from a project I finished recently that I want to write up because I keep seeing other people make the same mistake.

I built a research tool for a professional services firm. Took me about two weeks. I quoted €2,700 and they said yes immediately, no negotiation. Felt great at the time. A friend who runs a small agency looked at the numbers afterwards and told me I underpriced by 4-5x. The more I sat with it the more obvious it was that he was right. So here's the math I should have run before I sent the quote.

What I did wrong was straightforward. I priced based on what felt like a lot of money to me. I'm in Tbilisi. €2,700 for two weeks of work felt amazing. The problem is that "feels like a lot to me" has nothing to do with what the thing is worth to the buyer. It's just cost-plus pricing wearing a confidence costume.

Here's the math I should have run.

For a professional services firm this is unusually easy because they already think in billable hours. Their team was spending roughly 30 minutes per query doing manual research, somewhere between 8 and 15 queries per day across the team. Call it 50 hours a week of search time. Loaded hourly rate for that kind of specialist is conservatively €60-80, probably higher, but fine. That's €3-4k per week of labor going into manual document search.

The tool brought each query down to under a minute. Even if you assume only 70% of those recovered hours actually convert into billable client work (because not all of it does, some just gets absorbed back into the day), you're still looking at €2,000+ per week the system pays back. Annualized that's north of €100k.

Now put €2,700 next to that number. The build pays itself off in a week and a half. No wonder they didn't push back. I handed them a 35x first-year ROI and acted surprised when they signed.

The general lesson, the way I'd write it up for myself.

Step one is figuring out the annual value the buyer gets. Hours saved a week times their loaded hourly rate times 52. That's the only number that matters when pricing.

Step two is sanity-checking your quote against that number. Anywhere from 10-25% of year-one value is probably reasonable for a build, depending on how much risk you're taking on and how clean the deliverable is. Below that range and you're basically donating margin to the client.

The gut check that I think is the most useful part of all this: if the buyer doesn't push back at all, you priced too low. Some friction is the signal you're at the right number. Total absence of negotiation usually means you were never close to their walk-away price.

One nuance worth mentioning because it came up when I talked this through with people.

Underpricing your first big client isn't always a mistake. €2,700 sails past procurement, doesn't need approval, gets signed off on someone's personal authority. €12k probably triggers three meetings and a budget cycle and an "are we sure" email thread. When you need the case study more than the cash, the cheap-and-fast yes can be the right call. The actual mistake is letting that first price anchor what you quote everyone after.

The one-line version:

If you're building anything where the value to the buyer is measurable in time saved, and you're pricing off what feels like a lot to you, you're almost certainly leaving 3-5x on the table. Buyers aren't pricing your effort. They're pricing what your work gives them back.