What does an AI consultant do? A plain definition

An AI consultant audits how your business actually works, finds the tasks AI can usefully take over, then builds, governs, and hands over the systems that do it. The role spans five areas, audit, strategy, build, governance, and enablement, and the audit always comes first, because most buyers do not yet know which AI project is worth doing. AI consultancy and AI consulting for UK firms are two names for that same job.

That is the short answer. The rest of this post explains what each part looks like in practice, what services to expect, and whether you need a consultant at all if you already have an internal IT team.

What does an AI consultant actually do day to day?

Most of the work is unglamorous: watching how someone does their job, working out which parts are repetitive enough that AI can do them faster or more reliably, then setting up the systems to make that happen. The strategy deck is the small part. The hours go into discovery, integration, and the handover that decides whether anything sticks.

What separates a useful consultant from a slide vendor is that first part. Sitting with the team, watching how work moves, and asking what the bottlenecks are takes time and produces an uncomfortable answer surprisingly often: this process does not need AI, it needs fixing first. The consultants who skip discovery are the ones who ship the wrong thing.

What services do AI business consultants offer?

Most AI consultants offer some mix of five services. The table below is the honest version of what each one delivers and where the value sits. Weaker firms stop after strategy; the ones worth hiring also build the thing and leave you able to run it.

The split that matters is between firms that only advise and firms that also deliver. Advice without a build leaves you holding a roadmap and no closer to a working system. A build without enablement leaves you dependent on the firm that built it. The useful engagements cover the whole arc and write the consultant out at the end.

Do I need an AI consultant if I already have an internal IT team?

Often yes, but for a bounded period rather than forever. An internal IT team is built to keep existing systems running; an AI consultant brings something harder to acquire on a single project, which is pattern knowledge of which AI use cases pay off and which quietly fail. That knowledge is the thing you are renting, not the keyboard time.

The usual model is sequencing rather than choosing. A consultant scopes the opportunity, builds the first system, and documents it so your IT team can run and extend it once the engagement ends. Your team learns the system as it is built rather than inheriting it cold, which is the difference between capability that holds and a black box nobody internally understands. For the fuller decision on this, our guide on how to choose an AI consultancy and when to build in-house lays out the cost comparison and the conditions where in-house genuinely wins.

Where AI consulting helps, and where it does not

Not every process needs AI. We help UK businesses find the ones that do.

AI consulting helps when manual repetitive work is eating real hours every week, when your data is already reasonably structured, and when at least one person internally will champion the change. Those three conditions together are roughly the threshold where the spend starts paying back.

It does not help when the underlying process is broken in a way AI cannot fix, when your data lives in five disconnected systems with no shared identity, or when leadership is buying AI to look modern rather than to solve a named problem. A good consultant tells you that on the first call and saves you the engagement, which is also the cheapest way to spot one worth hiring.

How to tell a good AI consultant from a weak one

The clearest signal is what a consultancy shows you in the first call. A good one talks about a specific system they shipped, what the failure modes were, and what the rollout actually looked like. A weak one talks in frameworks and abstractions, because there is nothing concrete to point at.

Three things are worth verifying before signing anything.

• Ask to see a real working system they shipped, in production with a client actually using it, not a demo with sample data.
• Check they hold ISO 27001 or equivalent if you handle any commercial data, because AI work touches sensitive data by default.
• Ask who is going to do the build. Many large consultancies sub-contract delivery while keeping the senior names on the proposal; you want the team that pitches to be the team that builds.

If those three answers are clean, the rest of the engagement is usually a question of fit rather than capability.

How OpenKit runs an engagement

OpenKit works in three engagement shapes. Most businesses start with the first and decide later whether to continue.

• AI Audit. A scoped assessment at £10,000 excluding VAT: interviews and analysis, then the report, the plan and the costings. It ends with a prioritised roadmap and an estimated cost against each line, so you can decide what to do first.
• Embedded AI Lead. An ongoing engagement where an OpenKit team works alongside yours as your AI function, implementing the agreed priorities inside a reserved delivery allocation covering on-site work, remote implementation, meetings and support.
• Bespoke build. A major standalone application or a new system, for when an off-the-shelf tool is not the right answer. It needs its own scope, unlike the automations and connectors that sit inside an Embedded AI Lead engagement.

For how OpenKit positions consulting against bespoke development, our AI consulting service lays out the three shapes in more detail. For concrete examples of shipped work, the BAiSICS legal AI, MyPain healthcare AI, and Rubrical education AI case studies are the closest thing to “show me what you shipped”.