Earlier this year I put a proposal to a city for AI analytics work with a federated learning component. Technically sound. Genuinely useful. Priced at around one hundred thousand US dollars.

Their budget was twenty thousand.

My first explanation was that the market was not ready. That explanation lasted until I wrote it down and read it back. It is the most comfortable available story and it puts the failure outside the room.

What actually went wrong

The idea was not the problem. The funding unit was.

I had designed something that only one buyer could fund, then asked a buyer of that size to fund all of it. Public sector budgets in this part of the world are not small because the work is unimportant. They are small because they are annual, line-itemised, and set before anyone knew this problem existed. A department with twenty thousand dollars is not a department with a hundred thousand dollars that is being difficult. It is a department with twenty thousand dollars.

The correct move was available and I did not take it: build the instrument once, price access to it across twenty places, and keep the entry artefact free everywhere. Same total revenue, a fifth of the ask per buyer, and every participant makes the thing more valuable for the next one. Bespoke single-place builds have the opposite property. Each one starts from zero and teaches you nothing you can sell twice.

The part that is harder to admit

Quoting a hundred thousand dollars felt like seniority. It signalled that this was serious work by someone with a track record, and pricing it lower felt like conceding that.

That is vanity dressed as positioning. A price that nobody in your target market can pay is not a premium price. It is a decision not to sell.

What I do now

Three rules, all of them written down after this.

Price shared instruments, never bespoke single-place builds. Publish tiering by income band or arrivals, so a small destination and a large one can both say yes without negotiating. Keep something real and free in every market, because the free artefact is what makes the paid one legible.

I am publishing this because a body of work that only ever reports its successes is not evidence of anything. It is marketing with footnotes.