AI consultancy on a monthly retainer or a fixed-scope project: which suits your business?
When a business decides to use AI, the first question is what to build. The second, which weighs more on the budget, is how to buy it. Here are the two most common routes, when each one makes sense, and how to decide without regretting it later.
What each model is
Both models solve similar problems. What changes is how you pay, what gets agreed at the start, and what happens when priorities shift halfway through (because they always do).
Fixed-scope project
In a fixed-scope project, you and the supplier agree a scope at the start: what will be built, by when and for how much. The price is fixed or estimated against that scope, and it is usually paid in stages tied to milestones. The supplier delivers what was agreed and the project ends.
It works well when the problem is clear and is not going to change. The risk shows up when the business realises, mid-project, that the real problem was something else. Changing the scope almost always means a change request, with a new timeline and a new price.
Monthly retainer
On a retainer, you buy a set number of hours a month from a specialist. Those hours go to whatever matters most that month: mapping a process, connecting two tools, building an automation, training the team or fixing something that is already live.
At Kunavv, the retainer runs from 10 to 20 hours a month. You are not paying for a finished system. You pay for the work done, month by month, and you can change priorities without renegotiating a contract.
In practice, a good share of those hours turns into small automations. Before any code is written, each one is described in plain English so that everyone agrees on what it does. An illustrative example:
when a quote request lands in the shared inbox
then read the name, the job and the deadline
create the deal in the CRM
tell the account manager who owns it
if anything is missing, reply and ask the customerAutomations like this run on the tools the business already pays for, such as Microsoft 365 or Google Workspace, HubSpot, Xero or QuickBooks. They need no new system, and many take only a few hours to build.
The two side by side
The table sums up the differences that matter most to the decision.
| Criterion | Fixed-scope project | Monthly retainer |
|---|---|---|
| How you pay | A fixed or estimated price against a scope, usually in stages tied to milestones | A monthly fee for a set number of hours |
| Commitment | You commit to the scope and the price until the project ends | Month to month: you can stop at any time with one month’s notice |
| Agreed at the start | The full scope | The priorities for the first month |
| If priorities change | A change request: new scope, timeline and price | Next month’s hours go to the new priority |
| First delivery | After discovery and development | Can land in the first few weeks, using tools you already have |
| Biggest risk | Building something big for the wrong problem | Hours spread across tasks with no clear priority |
| Best for | A clear, stable problem that needs its own system | Problems still being discovered, and continuous improvement |
When a fixed-scope project makes sense
A fixed-scope project is not the villain. It is the right choice when:
- the problem is well defined and the requirements should not change over the next few months;
- the solution has to be a system of its own, because nothing on the market does the job;
- there is an in-house team, or a support contract, to look after the system once it is delivered;
- the expected gain pays for the investment, and someone has already done that sum with the company’s real numbers.
If the last point is missing, stop before you sign.
When a monthly retainer makes more sense
A retainer is usually the better place to start when:
- you know the team loses time to manual work, but you do not yet know which process to tackle first;
- the problems are many and small, spread across different areas such as customer enquiries, finance and sales;
- the business already pays for tools, such as spreadsheets, a CRM, accounting software and Microsoft 365, and wants more out of them before buying anything new;
- you want the team to learn to use AI day to day, not just receive a finished system.
The right question is rarely how much the system costs. It is which process hurts most right now, and what the simplest way to fix it is.
Vinícius de Borba, Kunavv
Five questions before you sign
Answer these five honestly. They settle most of the doubt.
- Can I describe the problem in one sentence? If not, you are still discovering the problem. At that stage, hours serve you better than a fixed scope.
- Will the problem be the same in six months? If your operation changes quickly, a scope signed today can be out of date before it is delivered.
- Who will look after it afterwards? Every system needs maintenance. If nobody in the company will own it, add that cost to the price of the project.
- Is my data organised? AI depends on data. If yours is scattered across loose spreadsheets and email threads, the first job is to organise it, not to build.
- What personal data will the supplier handle? If the work touches customer or staff data, the supplier will usually be processing it on your behalf, and UK GDPR requires a written contract setting out how. Ask for it before work starts, and ask which AI models will see the data.
What if the answer is to wait?
Sometimes the honest answer is neither, not yet. In one of our use cases, a law firm wanted to automate negotiations. The study took about 30 hours over three months and concluded that the firm’s data was not organised enough yet. The investment in a system was avoided.
An AI readiness assessment is how you reach that answer before you spend. It looks at data, processes, tools, governance and the team, and shows where the company stands today.
Summary
- Fixed-scope project: for a clear, stable problem that needs its own system and has someone to look after it afterwards.
- Monthly retainer: for problems still being discovered, many and small, solved step by step with what the business already has.
- Disorganised data: start there, whichever model you choose.
Last updated: