Fixed price, fixed scope: how we quote
We quote a fixed price against a fixed scope. Not an hourly rate, not a range that firms up later, not a number that moves once the work is underway. That choice is not a billing preference. It changes what both sides are incentivised to do for the length of the engagement.
Hours are the wrong unit
An hourly engagement pays the supplier for elapsed time, which means every efficiency the supplier finds costs the supplier money. That is a straightforwardly broken arrangement when the work is building systems whose entire purpose is to remove effort. It also gives the buyer nothing to approve: an estimate of hours is a guess about our internal process, and the buyer has no way to check it, no way to compare it, and no protection if it turns out to be wrong.
A fixed price transfers that risk to us, where it belongs. If the build takes longer than we thought, that is our estimate that was wrong, and our margin that absorbs it. The buyer approves one number against one described outcome, and the number does not move.
What a scope has to contain before we price it
A fixed price is only safe for either party if the scope is genuinely fixed, so the written statement of work carries four things: the deliverables, the timeline, the price, and — the one most proposals omit — what is explicitly not included. That last section is the one that prevents the argument. Anything outside it is a change order with its own price, agreed before it is built, rather than a quiet expansion nobody priced.
Revisions are bounded for the same reason. Two revision rounds per phase are included. “Until you are satisfied” sounds generous and is actually undefined, and an undefined commitment is one either side can be trapped by.
The productised entry point
Some work is repeatable enough to carry a single fixed number. A corporate website is one: base scope through deploy, two weeks, with languages, map modules and a content system quoted as separate lines rather than buried in the base. Agent setup is another: three weeks, one document or decision workflow put into production — ingestion, extraction, a human review queue, output — plus the evidence layer that records what ran, what it read, and what a human approved. Multi-agent orchestration, model training and unbounded integrations sit outside it, by design.
The number comes from a written schedule, not from a guess at what a given buyer will bear, and it is fixed in the statement of work before anyone builds anything. Larger engagements — AI strategy, audits and governance reviews, full agent systems, product builds — are scoped on a call, because their shape genuinely differs. We do not take on small projects.
When the budget is short, we cut scope
We do not discount. A discount says the original number was arbitrary, and it means the next buyer is paying for this one. When a budget does not reach the price, the honest move is to make the engagement smaller: fewer workflows, a narrower document set, one integration instead of three. The rate stays where it is and the scope comes down to meet the money, which leaves both the quote and the relationship intact.
The terms, plainly
Work starts on a 40% deposit, with the balance on delivery; larger engagements bill against milestones instead. Everything above is written into the statement of work before anyone builds anything. It takes one call to get there.
Related reading: what actually makes an AI agent trustworthy with money. Or see how we work.