Blog Details

For most SME AI projects, a fixed scope with a fixed price ships faster and protects the client better, because it forces both sides to agree on exactly what success means before any money is spent. Open-ended hours suit genuine research and shifting requirements, but they quietly transfer risk to the client, who pays for every detour. The honest answer is that good AI execution and delivery usually blends the two: a fixed-price discovery to remove the unknowns, then a fixed-scope build once the target is clear.

The two models, stated plainly

Every delivery contract is a way of splitting risk between you and your provider. The model you choose decides who pays when reality differs from the plan, and with AI, reality almost always differs from the plan.

Fixed scope, fixed price

You agree a defined deliverable for a defined amount. The provider carries the risk of overruns. In exchange, they scope carefully, pad for uncertainty, and push back on changes. You get budget certainty and a clear definition of done. You give up some flexibility.

Open-ended, time and materials

You pay for hours worked and adjust as you go. You carry the risk of overruns. In exchange, you get flexibility to change direction as you learn. You give up budget certainty, and you rely heavily on the provider's honesty about how those hours are spent.

Why fixed scope usually protects the client better

The instinct is that AI is uncertain, so you need open-ended hours to cope. That instinct is often wrong for an SME, and here is the uncomfortable reason.

Open-ended contracts reward the provider for taking longer. The slower the work, the bigger the invoice. Even honest teams drift when the meter is running and nobody agreed where the finish line is. For a small business without a technical manager watching every sprint, that is a dangerous incentive to sign up for.

• Fixed scope forces clarity up front. To price it, the provider must understand the problem, the data, and the definition of success. That discipline alone prevents a large share of failures.
• Fixed scope caps your downside. If the work is harder than expected, that is the provider's problem, not a surprise on your invoice.
• Fixed scope aligns incentives around finishing. The provider gets paid on delivery, so they are motivated to ship, not to linger.

The catch with fixed scope

Fixed scope only works when the scope can actually be defined. If the requirements are genuinely unknown, a provider forced to quote a fixed price will do one of two things: pad the price heavily to cover the risk, or lowball to win the deal and then fight you on every change. Neither serves you. This is the real limit of the model, and it is why the pure version is not always right.

Where open-ended hours genuinely fit

Time and materials is the honest choice in a few specific cases, and pretending otherwise would be selling you a false certainty.

• True research. When nobody knows if the thing is even possible, you cannot fix a scope around an unknown outcome.
• Ongoing evolution. Once a system is live and you want continuous improvement, a capped monthly arrangement often beats endless fixed-scope mini-projects.
• A capable client-side owner. If you have someone technical steering the work and checking the hours, the risk of open-ended drift drops sharply.

The blend that actually ships

The best AI delivery rarely picks one model for the whole engagement. It sequences them. The uncertainty in an AI project is heavily front-loaded, so you attack it first, cheaply, and only commit to a fixed build once the fog has cleared.

Phase one: a fixed-price discovery

Buy a short, fixed-price discovery whose deliverable is clarity: the specific problem, the state of the data, the recommended approach, and a fixed-scope proposal for the build. This is cheap, time-boxed, and it converts unknowns into a defined target. If the discovery reveals the project should not proceed, you have saved yourself the entire build budget.

Phase two: a fixed-scope build

With the unknowns removed, the build can now be scoped and priced with confidence. You get budget certainty exactly when the stakes are highest, and the provider can commit because they are no longer quoting blind. Structured AI execution and delivery is built around this sequence, turning a scary open-ended commitment into two controlled steps.

How to choose in one paragraph

If you can describe the outcome you want in a sentence and the data plausibly exists, push for fixed scope and make the provider earn it. If the outcome is a genuine unknown, buy a small fixed-price discovery before you commit to anything open-ended. Reserve pure time and materials for real research or for improving a system that is already live. And whatever the model, insist on a defined success metric, because a contract without one protects nobody.

A worked example of the blend

Picture a distributor who wants AI to answer routine customer emails. Under a pure open-ended contract, they would sign up for an unknown number of hours and hope. Under a pure fixed-scope contract, a provider would either pad heavily to cover the unknowns or lowball and fight over changes. Neither serves the distributor well, because at the start nobody actually knows how messy the email history is or whether the answers can be trusted.

The blend fixes this. First, a short fixed-price discovery examines the email archive, tests whether a model can draft reliable replies, and returns a clear recommendation with a fixed-scope proposal. That step is cheap and time-boxed. If it turns out the data is too thin, the distributor has spent a small sum to avoid a large mistake. If it looks promising, the build is now a known quantity that can be priced with confidence. The uncertainty was paid for once, cheaply, and then removed.

Red flags in each contract

Whichever model you choose, certain warning signs mean you are about to be exposed. Learn to spot them before you sign.

• Fixed scope with no discovery. A provider who quotes a firm price for a vague AI project without investigating your data is either padding heavily or planning to win the fight over changes later.
• Open-ended with no cap. Time and materials without a ceiling and regular checkpoints is a blank cheque. Insist on a cap and a review rhythm.
• No success metric in either. A contract that does not define what good looks like protects the provider, not you, regardless of the pricing model.
• Vague deliverables. If you cannot tell from the contract what you will actually receive, the model does not matter, because you cannot hold anyone to it.

What belongs in every contract, whatever the model

Some protections are model-agnostic. A clear definition of done, an agreed success metric, a named owner on both sides, and a handover clause that leaves the code and documentation with you. These matter as much as the pricing structure, and they are where an SME without a technical manager is most often let down. The delivery model decides who carries the risk of overruns. These clauses decide whether you are left with an asset or a dependency.

The role of trust

Underneath the whole debate sits a simple truth. Open-ended contracts require far more trust than fixed-scope ones, because you are paying for effort you cannot fully verify. If you do not yet know a provider, a fixed-price discovery is also a low-risk way to test them. You learn how they work, how honestly they communicate, and whether they deliver what they promised, all before you commit real budget. Earning the right to an open-ended relationship is something a good provider is happy to do.

Match the model to the moment

The mistake most SMEs make is choosing a delivery model out of habit rather than fit. They default to open-ended because it feels flexible, or to fixed scope because it feels safe, without asking where their project actually sits on the uncertainty curve. The honest question is simple. How much do you genuinely know about the problem and the data right now?

If the answer is a lot, fixed scope rewards that clarity and protects your budget. If the answer is very little, do not paper over the gap with an open-ended blank cheque. Buy the clarity first, cheaply, then convert it into a fixed commitment. The model should follow the state of your knowledge, and that state changes as the project moves. The best engagements shift models on purpose as the fog clears, rather than locking into one and hoping it fits the whole journey.

The bottom line

Fixed scope is not the safe-but-slow option and open-ended is not the flexible-and-smart option. For most SMEs, fixed scope ships faster and protects you better because it forces clarity and caps your downside. Open-ended hours have their place, but they quietly hand you the risk. The strongest path is to buy clarity first at a fixed price, then buy the build at a fixed scope. Certainty where it matters, flexibility where it helps.

Unsure which model your project needs? Request a free AI opportunity assessment and we will recommend the delivery structure that fits, with no obligation.