Two weeks to know which AI system pays you back. Twelve to be using it.

The assessment ends with the right case chosen, priced and ready to enter development — not with a report to file away.

Before building anything, you need to know what is worth building. The assessment separates the processes where AI returns time and money from those where it only adds a layer of uncertainty to something that already worked.

The problem with starting from the technology

Most AI projects that fail do not fail at the modelling stage. They fail because the model was chosen first and a problem for it to solve was found afterwards. The result is a pilot that impresses in a demo and is irrelevant in operations.

We reverse the order. We look at the processes that already exist, measure where time is lost and where errors cost money, and only then ask whether there is anything here a model does better than a simple rule.

How we work

• Discovery. Sessions with the people who run the process, not only those who describe it. We map volume, time per task, decision points and what goes wrong today.
• Data inventory. What data exists, where it lives, in what state and who may see it. This is where most use cases die — and it is far better for them to die here than six months later.
• Prioritisation. Each case gets an effort estimate, expected return, associated risk and dependencies. We rank by return over effort, not by how impressive it looks.
• Recommendation. We say what to do first, what to postpone and what not to do at all.

What you get

• A report with the identified use cases, prioritised, each with an effort and return estimate.
• A map of the data required and what is missing to get it into shape.
• The risks of each case — quality, privacy and vendor dependency — and how to contain them.
• A proposed first case to put into production, with the success criterion defined before work starts.

From assessment to a system built

The assessment does not end in a report: it ends with the first case chosen, priced and ready to go into development. We are a team that builds — what comes out of this becomes code, not slides. On most projects the first system is in production six to twelve weeks later.

We develop on Claude, Google Gemini and OCI Generative AI, integrated into the systems you already run. Where the right answer is deterministic automation rather than a model, we build that instead — cheaper to run, and from the same team.

Frequently asked questions

How long does an assessment take?

Two to four weeks, depending on the number of processes under review and how available people are for the discovery sessions.

Do we need our data organised before starting?

No. Assessing the state of the data is part of the work. What often happens is that the assessment shows the right first investment is not in AI at all, it is in organising the data.

Does the assessment commit us to working with you?

No. The report is yours and any team can execute it. If you decide to go with another supplier, you have everything you need to brief them.

← All services