How to choose an AI consultancy in Wales

Who wrote this guide

InnovAItive Solutions is an AI consultancy in Cardiff, so we are one of the options you might compare. We have written this to help you choose well, not to rank providers. Where another type of provider is the better answer, we say so and link to it.

The four questions that separate a consultancy from a sales process

1. What will you look at first? A good answer starts with your work and the tools you already pay for. Be wary of a proposal that names a product before anyone has seen how your business runs.
2. What does the first step cost, and what do we get? Ask for a fixed price, a fixed scope and a named deliverable, such as a written report and a roadmap. Open-ended day rates make it hard to stop.
3. Who checks the AI's output? Ask how a person reviews anything that reaches a customer, a payment or a regulated decision, and where your data will be processed. UK GDPR applies from the first pilot.
4. What will we own when you leave? Prompts, workflows, documentation, usage rules and trained people should stay with you. Ask who owns the code and the accounts.

What each delivery model costs

Prices vary, so ask every provider for theirs in writing. These are the common models and how they are usually charged.

• Community and publicly funded support. Often free or subsidised awareness sessions and advice. Business Wales lists current public support, which changes over time.
• Fixed-price discovery. A set day or sprint with a fixed fee and a written output. Ours, the AI and Copilot Discovery Day, is £950 + VAT. Other UK consultancies publish their own discovery prices, some of them higher, so compare scope as well as price.
• Day-rate consultancy. Charged per consultant day. Ask for the day rate, the expected number of days and what happens if it runs over.
• Fixed-scope build or pilot. A quoted price for one automation or AI agent, usually after discovery. Ask what is included in support after go-live.
• Retained or embedded support. A monthly fee for an agreed amount of time, such as our Personal Forward Deployed AI Engineer. Ask for the notice period.
• An in-house hire. A salary plus on-costs, recruitment time and management. Best value when AI work will keep one person busy for years.

What a readiness assessment should contain

Whoever you choose, the first piece of work should leave you with these, in writing.

• A map of the tools you already pay for, what overlaps and what is unused.
• A view of your data: where it lives, who can see it, and what must stay out of AI tools.
• A short list of opportunities scored by value, effort and risk, including the ones where the answer is "not AI".
• Clear usage rules for your team and a named owner for each tool.
• A 90-day plan with simple measures, so you can tell whether the first pilot worked.

When a different provider is the better answer

• A community or publicly funded body if you want free, general awareness before spending anything.
• A Microsoft partner if you mainly need licences, a tenant-wide roll-out or ongoing Microsoft 365 administration. Microsoft publishes a partner directory.
• A generalist digital agency if AI is one small part of a wider website, marketing or app project they already run for you.
• An in-house hire if AI is core to your product, or will keep a person busy full time for years.
• A specialist data science firm or a university if you need model training or research rather than adoption of existing tools.
• An AI consultancy like ours if you want an independent view of your tools and data, a fixed-price first step, and a team that can also build and pilot what it recommends.

Simple checks before you sign

• Look the company up on Companies House: incorporation date, registered office and filed accounts.
• Ask for a client you can speak to, or an honest answer if they cannot share one yet.
• Ask which published results are their own client work and which are illustrations or other people's research.
• Read the contract for data processing terms, ownership and notice periods.

Questions we are often asked