AI Consulting and Strategy from Use Case to Costed Roadmap

Teams that have several possible AI projects and need to choose between them on evidence: operations and product leaders, CTOs, and founders deciding which AI feature comes first. If you already know exactly what to build, you may be better starting with a scoped pilot or an AI MVP.

A short, throwaway test of one idea on a sample of your real inputs, using off-the-shelf models. It answers practical questions — how often is the output good enough, where does it fail, what would each task cost — before anyone commits to a build. Spikes produce evidence, not production software.

A focused engagement runs as a discovery sprint, from £2,000 as a fixed fee agreed before we start. Model fees for the feasibility spikes are paid to the provider and agreed with you in advance. Wider reviews across several departments are quoted as a fixed-price project.

We gather as many ideas as your teams have, and score them all on paper. Only the strongest two or three go through feasibility spikes, because testing on real samples is where the time goes — and where the useful evidence comes from.

A sponsor who can make the final decision, the people who do each shortlisted task today, and someone who knows the systems and data involved. We work remotely, over video calls and a shared channel, with weekly written updates during UK business hours.

By the shape of the task. Reading, drafting and answering questions from text suit language models; predicting a number or a risk from historical records usually suits classic machine learning; fixed, well-understood logic suits plain rules. Many good solutions combine them, and the spikes show which works on your data.

The shortlisted use cases in order, with the reasoning for that order; the shared groundwork, such as data access or search, that several use cases depend on; a cost range to build and to run each one; the measure that defines success; and the people who need to sign off.

Yes. For a pilot, we look at how it was tested, what it costs to run and what is stopping it going live. For a supplier’s proposal, we test the key claims on a sample of your own data where possible and set out the questions to ask before you sign.

Then that is the recommendation, with the reasons written down. A clear “not yet” or “buy this instead” saves far more than it costs. Often the better first step turns out to be a process fix, a report or a simple automation, and we will say so.

It flags what your advisers need to look at: personal data in each use case, where a person must approve decisions about individuals, and what should be logged. UK GDPR and ICO guidance apply to most use cases involving personal data, and the EU AI Act may apply if you sell into the EU. Your DPO and legal team make the call — we do not give legal advice.