AI projects are usually priced in one of three ways: fixed-price (a set fee for a defined scope and outcome), time-and-materials (you pay by the hour for however long the work takes), or staff augmentation (you rent specialists by the day or month). Fixed-price puts the delivery risk on the vendor and gives you a known cost up front; time-and-materials and staff augmentation put the risk on you, because the bill grows with the hours. For a discrete AI build with a clear goal, fixed-scope, fixed-price is almost always the safer choice for the buyer.
The three ways an AI project gets priced
Before you compare vendors, it helps to know that almost every AI engagement is sold using one of three commercial models. They sound similar in a sales call, but they hand the financial risk to very different parties.
- Fixed-price (fixed-scope): You and the vendor agree on a specific deliverable and a single price. If it takes longer than expected, that is the vendor's problem, not yours. In Dutch this is the opposite of *nacalculatie* — the number is fixed before work starts.
- Time-and-materials (T&M): You pay an hourly or daily rate for as long as the work runs. This is the classic *uurtje-factuurtje* / *nacalculatie* model. The scope can flex, but so can the invoice — there is no contractual ceiling unless you negotiate one.
- Staff augmentation: You rent one or more specialists (a data scientist, an ML engineer) by the day or month and direct them yourself. You are effectively hiring temporary headcount. The vendor owes you a person, not an outcome.
The difference that matters is not the rate card. It is who is accountable for the result, and who absorbs the cost when reality turns out messier than the proposal.
How fixed-price works — and what you give up
Under fixed-price, the vendor commits to a defined outcome for a defined fee. To do that responsibly they have to invest real effort up front: understand your data, scope the problem tightly, and price in their own uncertainty. That scoping work is a feature, not overhead — it forces hard questions to surface before a contract is signed rather than three months into an open-ended build.
The clear win for you is predictability. You can take a single number to a budget holder, compare it against the expected return, and decide. There is no meter running. If the model is harder to train than the vendor assumed, the fee does not move.
What you give up is flexibility mid-flight. Because the price is anchored to a defined scope, genuinely new requirements are handled through a change request, not absorbed silently. That is not a trap — it is the mechanism that keeps the original number honest. A good vendor will tell you plainly when something is in scope and when it is a change, and let you decide.
Fixed-price works best when the goal is concrete: automate this document workflow, build a retrieval system over these knowledge bases, ship a proof of concept that proves one hypothesis. For genuinely exploratory research with no defined endpoint, a fixed price is harder to set fairly — and an honest vendor will say so.
How time-and-materials works — and why the buyer carries the risk
Time-and-materials is the default model for a lot of consultancies because it is low-risk for *them*. They bill what they work, so they are never out of pocket. The flip side is that every hour of inefficiency, every false start, and every unclear requirement lands on your invoice.
On AI projects specifically, that risk is amplified by a few things that are genuinely hard to estimate: data quality is rarely as good as anyone hopes, the first model approach often does not work, and "good enough" accuracy can be a moving target. Under fixed-price, the vendor priced that uncertainty in and owns it. Under T&M, you are funding the discovery in real time, with no obligation on the vendor to be fast about it.
The deeper problem is scope creep with no defined endpoint. When billing is open-ended, there is little structural pressure to declare the project done. "One more iteration" is always available, and it is always billable. Projects drift, budgets quietly double, and the conversation shifts from "did we get the outcome" to "how many hours have we burned."
None of this means T&M is dishonest — it is a legitimate model, and for a long-running internal capability with a trusted partner it can be the right one. But for a discrete project, you should go in clear-eyed that you are the one holding the financial risk, and you should at minimum demand a not-to-exceed cap and milestone gates. If you want the mechanics of estimating a bounded AI build, our note on what an AI proof of concept costs walks through it.
Staff augmentation: renting people, not buying outcomes
Staff augmentation is a third model that is easy to confuse with consultancy but is fundamentally different. You are not buying a result — you are buying a person's time and slotting them into your own team and management structure.
This can make sense if you already have a strong internal AI lead, a clear plan, and a temporary gap in hands. You keep full control of the architecture and the roadmap, and you can scale the team up or down. Nearshore and offshore providers compete hard here on day rates.
The risks are the ones you would expect from temporary headcount. Accountability for the outcome stays with you, not the vendor — if the project fails, the vendor still delivered "a competent engineer for six months," which is all they promised. You also carry the management overhead, the onboarding cost, and the knowledge that walks out the door when the contract ends. And because day rates feel modest line by line, the total can quietly exceed what a scoped project would have cost.
Crux Digits is deliberately not a staff-augmentation shop. We do not rent out bodies by the day or run dedicated-team arrangements. We take responsibility for a defined outcome and deliver it — which is a different commercial promise entirely.
Why Crux Digits prices fixed-scope, fixed-price
We chose fixed-scope, fixed-price because it aligns our incentives with yours. When the price is fixed, we are paid to be *efficient and effective*, not merely busy. If we are slow or take a wrong turn, that is our cost to absorb, not yours. That is exactly the incentive you want pointing at a vendor.
It also means you know the cost and the outcome before you commit. Our engagements are published, not negotiated case by case: an AI Audit & Strategy at a fixed EUR 2,500, a Proof of Concept at a fixed EUR 20,000, and Production Launch from EUR 50,000. You can see the full ladder on our pricing page. No meter, no surprise true-up at the end of the quarter.
Being honest about this means being honest about its limits too. Fixed-price requires disciplined scoping, so we spend real time defining the problem before quoting — and we will tell you when a request is genuinely out of scope rather than letting it inflate quietly. We think that transparency is the point. You can read more about how we work across AI consulting and AI automation, or how we structure an engagement in our guide to running an AI pilot.
How to choose the right model for your project
There is no single correct answer — the right model depends on how well-defined your goal is and how much delivery risk you want to carry yourself.
- Choose fixed-price when the outcome is clear and you want budget certainty: a specific automation, a proof of concept, a first production system. You trade some mid-project flexibility for a known cost and a vendor on the hook for delivery.
- Consider time-and-materials when the work is genuinely open-ended research, or an ongoing partnership where requirements evolve weekly and you trust the partner. If you go this route, insist on a cost cap and milestone reviews.
- Consider staff augmentation only when you have strong internal leadership and a clear plan, and you simply need extra qualified hands for a defined period. Remember the outcome is still yours to own.
A useful test: can you write down, in one or two sentences, what "done and successful" looks like? If yes, fixed-price will protect you. If you genuinely cannot — and that is a real and acceptable situation in early research — then a capped T&M arrangement with frequent checkpoints is more honest than a fixed price built on guesses.
What to ask a vendor before you sign
Whatever model a vendor proposes, a few direct questions will tell you fast whether the commercials are designed to protect you or just to protect them.
- Who carries the risk if it takes longer than planned? Under fixed-price the answer is the vendor; under T&M it is you. Make sure the answer matches the price tag.
- What exactly is in scope, and what counts as a change? A vendor who can answer this crisply has done the scoping work. A vague answer is a warning sign whatever the model.
- What does "done" mean, and how is success measured? If there is no defined endpoint and no success metric, an hourly engagement has no natural off-ramp.
- Is there a not-to-exceed cap? For any T&M arrangement, no cap means open-ended liability for you.
- Do I own the code, the models, and the data pipelines at the end? This matters most with staff augmentation and ongoing partnerships — make ownership explicit.
- What happens after handover? Support, retraining, and maintenance are real costs. Find out whether they are included, fixed, or another open meter.
If a vendor cannot give you straight answers to these, the pricing model is the least of your concerns. If you want a second opinion on a proposal you have received, a short conversation with us is free — we will tell you honestly whether the commercials look fair, even if the work is not for us.
The bottom line
Pricing is not a billing detail — it is a statement about who is responsible for the result. Fixed-price says the vendor will own the outcome and the cost of getting there. Time-and-materials and staff augmentation, in their honest forms, say you will. For most companies commissioning a discrete AI build, that distinction is the single most important thing in the contract.
Our advice, even setting aside that it is how we work, is to favour a defined scope and a defined price whenever your goal is clear enough to write down. It keeps the vendor focused on shipping something that works, and it keeps your budget under your control. If your goal is not yet clear, the first paid step should be the cheap one that makes it clear — a small, fixed-price audit or proof of concept — not an open-ended hourly engagement you cannot see the end of.
If you would like to talk through how a fixed-scope engagement would map to your specific project, reach out for a free, no-obligation conversation. We will be straight with you about cost, timeline, and whether the work is even worth doing.
Frequently asked questions
What is the difference between fixed-price and time-and-materials in AI projects?
Fixed-price means you agree on one fee for a defined scope and outcome, and the vendor absorbs any overrun. Time-and-materials (uurtje-factuurtje / nacalculatie) means you pay by the hour for however long the work takes, so the cost is open-ended and the risk sits with you. Fixed-price gives you budget certainty; time-and-materials gives the vendor a guaranteed margin.
Why is hourly billing risky for AI projects specifically?
AI projects carry hard-to-estimate uncertainty — data quality, model performance, and shifting accuracy targets all extend the work. Under hourly billing you fund all of that discovery in real time, with no contractual endpoint, so scope creep and a doubling budget are common. There is little structural pressure to declare the project finished when every extra iteration is billable.
What is staff augmentation and how is it different from a fixed-price project?
Staff augmentation means renting specialists by the day or month and managing them inside your own team — you buy people's time, not a result. With a fixed-price project, the vendor commits to delivering a defined outcome for a set fee and is accountable for it. With staff augmentation, accountability for the outcome stays with you. Crux Digits delivers fixed-scope outcomes and does not do staff augmentation.
How much does an AI project cost in the Netherlands?
It depends on scope, but transparent fixed-price vendors publish their numbers. Crux Digits prices an AI Audit & Strategy at a fixed EUR 2,500, a Proof of Concept at a fixed EUR 20,000, and Production Launch from EUR 50,000. Knowing the number before you commit is one of the main advantages of a fixed-scope model over an open-ended hourly one.
How do I choose between fixed-price, time-and-materials, and staff augmentation?
Ask whether you can write down what "done and successful" looks like. If you can, fixed-price protects you with a known cost and a vendor on the hook for delivery. If the work is genuinely open-ended research, use capped time-and-materials with milestone reviews. Choose staff augmentation only when you have strong internal leadership and just need extra hands for a defined period.
What should I ask an AI vendor before signing a contract?
Ask who carries the risk if the work runs long, what is in scope versus a billable change, how success is measured, whether there is a not-to-exceed cap, who owns the code and models at the end, and what post-handover support costs. Crisp answers signal a vendor who has scoped the work properly. Vague answers are a warning sign regardless of the pricing model.