September 7, 2026 · 6 min read
Fixed price AI development: how the number gets fixed
By Sahan, co-founder, systems and delivery
Fixed price AI development means the scope is written, agreed and costed before any code exists, and the number does not move while the work runs. The buyer carries no estimate risk. The builder carries all of it, which is only survivable if the scoping is done honestly before the price is quoted.
Our published rates as of July 2026 work this way: $1,900 for an operations audit over 10 working days, from $9,500 for a build sprint over 4 to 6 weeks, $2,900 a month for a fractional officer, cancel monthly.
Fixed price means the estimate risk sits with the builder
In a fixed price engagement the client pays an agreed number for an agreed outcome. If the build takes twice the hours we thought, that is our problem, not a change order. In hourly billing the reverse is true: a bad estimate becomes the client’s invoice, and nobody finds out until the third month.
That is the whole difference. Everything else people say about the two models is downstream of who eats a bad estimate.
Hourly billing survives because most AI projects do not finish
Gartner predicted on 25 June 2025 that over 40% of agentic AI projects will be canceled by the end of 2027, naming escalating costs, unclear business value and inadequate risk controls as the causes. The same release estimated that only around 130 of the thousands of vendors claiming agentic products are real.
MIT’s Project NANDA reached a harder number in July 2025. Its report “The GenAI Divide: State of AI in Business 2025”, built from 52 executive interviews, 153 survey responses and 300 public deployments, found that 95% of enterprise generative AI pilots produced no measurable profit and loss impact.
Cost control is where it shows up first. McKinsey’s Enterprise AI FinOps survey of 75 enterprises, fielded in May 2026, found 93% had already exceeded their AI budgets, with roughly 60% of agentic AI cost sitting in response refinement, a line item most budget models never had.
Read those three together and hourly billing looks less like a pricing preference and more like a hedge. If a large share of this work will not finish, billing by the hour is the only way to get paid for the part that does not.
There is a second cost to hourly work that rarely appears in the comparison. It removes the deadline. A fixed number and a fixed date fail loudly if they slip, so both sides watch them. An hourly engagement fails quietly, one extra fortnight at a time, and the first honest conversation about scope happens after the money is spent.
What a fixed price forces the builder to do before quoting
To quote a number you have to know what done means. That means writing down the inputs, the outputs, the edge cases and the acceptance test before the first commit, not discovering them in week five.
Where the process is not understood well enough to write down, the honest answer is to charge for the writing down. That is what the $1,900 audit is: 10 working days, a full refund if we find nothing worth building, and the fee credited against the build if one follows. It exists so that the sprint price can be a real number rather than a hopeful one.
We can quote this way because we have built the same shapes before, on our own group first. The agent fleet page carries the dated version: 41 registered agents, 9 live in production, 5,450 runs with zero failures across 13 days, as of July 2026. A tomte, our word for one production agent with one defined job, has a scope narrow enough to estimate. That narrowness is the pricing mechanism, not a stylistic choice.
The written scope is also the acceptance test. Acceptance is measured against that document and nothing else, which is what stops the last two weeks of a project turning into a negotiation about what was implied in a meeting nobody minuted. We do not hold those meetings, so there is nothing to remember differently.
Fixed price versus hourly, from the buyer’s seat
| Question | Fixed price | Hourly or time and materials |
|---|---|---|
| Who pays for a bad estimate | The builder | The client |
| When the final number is known | Before the work starts | After the work ends |
| What “done” means | Written acceptance test | Whenever the budget stops |
| What a change costs | A new order, quoted before it starts | Absorbed silently into the run rate |
| Incentive on speed | Finish early, keep the margin | Every extra week is revenue |
| What a reseller can put in a proposal | A firm cost line on Monday | A range and a caveat |
The last row is the one agencies feel. You cannot quote a client a fixed number against a subcontractor who bills you by the hour, because you would be underwriting somebody else’s estimate with your own margin.
Change means a new order, not a surprise invoice
The discipline only holds if the change rule is stated up front. Ours: change after the scope is signed becomes a new order, quoted and agreed before anyone works on it. That reads harsh until you notice the alternative, which is a change absorbed quietly at the cost of the delivery date you were promised.
It also puts real weight on the scoping conversation. When a late change costs a fresh quote, both sides think harder in week one, which is exactly when thinking is cheap.
Why resellers need this more than direct buyers
Most of our work is direct with the company that will run the system, and that stays the core of the business. Agencies are a second route into the same delivery, and fixed pricing is what makes that route work at all.
Promethean Research’s 2026 State of Digital Services, fielded in February 2026 across 119 digital agency owners and managers, found 34% had implemented AI across the business and another 28% were actively implementing it. Those agencies are being asked for AI builds now, and a proposal needs a cost line, not an hourly rate with an asterisk.
We have written the arithmetic out before in the margin math of reselling AI development, and the delivery mechanics in white label AI automation. Both depend on the subcontract price being fixed first.
What a fixed price does not cover
It does not cover exploratory work with no defined outcome. If nobody can say what done looks like, the audit is the right purchase and the build is not.
It does not cover us operating your systems forever. A sprint ships to your own accounts, with two weeks of written post-handover support, and then it is yours. Ongoing ownership is a separate monthly arrangement.
It does not cover unlimited concurrency. One sprint runs at a time, which is part of why the dates hold.
And it does not make the work cheap in every case. In 2026 the market prices audits at $5k to $15k and single builds at $3k to $15k, so our numbers sit inside the market rather than under it. What changes is not the price, it is who carries the risk of the estimate being wrong.
Start
Describe the process you want built at /start. You get a written fixed price scope back: inputs, outputs, edge cases, what done means, and the number. Written reply within one business day. No call required, ever. If the scope is not clear enough to price, we will say so and quote the build sprint only once it is.