The contract model matters less for the money than for what it does to behavior once the project is running. Both common models create predictable incentives, both fail in predictable ways, and choosing between them is really a question about how well the work is understood before it starts.
What fixed price actually does
A fixed price transfers the risk of the estimate being wrong from you to the supplier. That is its entire purpose and it is a real benefit, priced into the number.
It also does three other things. It makes every change a commercial event, which slows down changes that would improve the product. It rewards a narrow reading of the scope document once the work turns out harder than expected. And it pushes the supplier to build the least expensive thing that satisfies the wording, since anything more comes out of their margin.
None of that is bad faith. It is what the structure asks of a rational supplier. A fixed price works well when the scope is genuinely well understood, and it turns adversarial exactly when the work turns out to be different from what was described, which is the moment you most need cooperation.
What time and materials actually does
Time and materials keeps the risk with you and removes the friction from change. Changing direction costs nothing contractually, which matters a lot when the right answer is being discovered as you go.
Its failure mode is the absence of a forcing function. Without a fixed end, scope grows quietly, decisions get deferred because deferring is free, and the project can run for a long time while everyone reports progress honestly. The budget becomes the only constraint, and it constrains from behind rather than in front.
Time and materials works when trust is established and the work is genuinely exploratory. It works poorly as the model for a first engagement with a supplier you have not seen deliver anything.
The variable that decides it
Neither model is better in general. The question is how much of the work is understood.
If the requirements are stable, the integrations are known, the data is understood and someone can describe the finished thing precisely, fixed price is appropriate and you should expect a firm to offer it.
If the project involves discovering what to build, integrating with a system nobody has looked inside yet, or migrating data of unknown quality, a fixed price is either padded heavily or will be defended narrowly later. Both outcomes cost you.
The problem is that most projects are a mixture, which is why the mixture should be priced rather than the whole.
Fixing the scope before the price
The structure that avoids both failure modes is to pay separately for the work of understanding, then fix the price on what is now understood.
A paid discovery produces a scope, an exclusion list, an architecture sketch, the assumptions and the risks. With that in hand, a fixed price means something, because the thing being priced has been examined rather than described.
That is how we work. Discovery runs five to ten days and ends with a statement of work carrying fixed pricing, assumptions, exclusions and risks. Scope is fixed before price, in that order. Discovery starts at $3,000 and the fee is credited toward the build.
The second part of the structure is phasing. Each phase is separately scoped and separately priced, and each ends with something working. That keeps every individual commitment small, which matters given that the Standish Group's CHAOS research puts small project success at roughly 90 percent against under 10 percent for large ones. It also gives both sides a natural point to renegotiate when reality differs from the plan, without either party having to breach anything.
Reading a fixed price quote
Ask what has been assumed. Every fixed price rests on assumptions, and a quote that does not list them has either hidden them or not made them consciously. The assumptions are where the disagreements will come from.
Ask what happens when an assumption turns out to be false. A named change process is fine. Silence means the conversation happens later, under pressure, with money already spent.
Look at whether the exclusions are specific. "Third-party integrations not listed above" is real. "Anything outside the agreed scope" is not, because it just restates the problem.
Compare quotes on assumptions rather than totals. Two fixed prices that differ by a factor of three are almost always pricing different projects, and the assumption lists will show you exactly where they diverged.
Reading a time and materials arrangement
Ask for a cap, or at least a review point. Time and materials without a stopping mechanism is an open commitment, and suppliers who are comfortable with a review point are the ones who expect to be doing well at it.
Ask what gets reported and how often. Hours consumed is not progress. What shipped to a real environment is progress.
Ask who decides priority. If it is not clearly you, time and materials becomes a subscription to a team rather than a route to a product.
Frequently asked questions
Which is cheaper overall?
Fixed price is usually higher on paper because it includes the cost of carrying risk. Time and materials is often higher in the end because nothing stops the scope growing. Neither is reliably cheaper, which is why the choice should be made on fit rather than on price.
Is fixed price possible for AI projects?
For the integration work, yes, once discovery has established what data is accessible and in what state. For work where the output quality is the deliverable, a fixed price on a defined evaluation target is realistic, while a fixed price on an undefined quality bar is not.
What about fixed budget with flexible scope?
This is often the best of the three. The budget is certain, the scope adjusts as you learn, and the incentive is to spend the budget on whatever turns out to matter most. It requires enough trust to work, which is why it usually follows a first engagement rather than starting one.
Should the first project with a new supplier be fixed price?
Usually yes, and small. A contained fixed-price piece lets you see how the firm behaves when something turns out harder than expected, which is the thing you actually need to learn.
How much should discovery cost?
Enough that it is real work rather than a sales document. Ours starts at $3,000 and is credited toward the build. Free discovery is a sales activity and tends to produce a scope shaped to win the work.
Where to start
If you are about to sign a fixed price for a project nobody has examined properly, the cheaper order is to examine it first. Our discovery produces the scope, the exclusions and the risks, then the fixed price follows.