How to choose an AI consultancy

Before comparing firms, be clear about the kind of help you need. Choosing an AI consultancy is easier when you know whether you want advice on where AI fits, someone to build and integrate a system, help automating a specific process, or training for your team. Some firms do one of these well; fewer do all four with the same care.

Write a short paragraph describing the problem, the people involved and what a good outcome would look like in six months. If you cannot yet describe the problem, that is a sign you need advice before delivery, and a good consultancy will start there.

A first conversation should tell you as much about how a firm thinks as what it sells. These questions are a useful test:

• How would you decide whether AI is the right answer here? Listen for an approach that starts with the workflow and data, and includes the possibility of saying no.
• What would the first piece of work be, and what would it produce? A clear, bounded first step is a good sign. A long programme before anyone has looked at your work is not.
• How do you handle our data? Expect specifics about tools, data terms, access, retention and UK GDPR, not reassurance.
• Who checks the output, and what happens when the system is wrong? Dependable systems are designed around review and failure, not just the happy path.
• What will our team be able to do without you afterwards? Good work leaves capability behind rather than dependence.
• Can we see relevant work? Ask for examples you can verify, such as live websites and systems, and a conversation about similar engagements.

Some patterns should make you cautious. Be wary of a firm that proposes a specific tool before understanding your work, promises headcount savings in the first meeting, or cannot explain how data will be protected. The same applies to anyone who treats governance as paperwork to be added later, or whose proposal has no point at which you can stop if the case is not there.

Be equally careful with claims that cannot be checked. Percentage improvements without context, unnamed case studies and certainty about results are easy to write. Ask how a figure was measured and what it would look like in an organisation like yours.

For most organisations, a good first engagement is short and specific. It usually includes a conversation with the people who do the work, a look at a small number of real workflows and the data they rely on, and a written recommendation that ranks options by value, risk and effort. It should be clear about what not to do yet.

From there, work can move to a focused pilot, a policy and training, or an implementation plan. Our own approach follows that shape: Orientation, Assessment, then Design, Build and Capability only where the case is there. You can read more about how we approach AI consulting, and our guide to building or buying AI helps frame the investment decision.

Much AI work can be done remotely, but being able to meet in person helps when a process spans a site, several teams or a customer-facing operation that is easier to understand in context. For organisations in the North West, a local adviser can make workshops and walk-throughs straightforward. We work on site across Widnes, Liverpool, Manchester, Chester, Wirral and beyond, and remotely across the UK.