Blog

What does an AI consultant do, and do you need one?

Benjamin · · 4 min read

An AI consultant helps a business decide where AI will genuinely help and how to put it in place. Here is what the work involves, the different kinds of AI consulting, and questions to ask before hiring one.

An AI consultant helps a business work out where AI will genuinely help, where it won't, and how to put it in place safely. The good ones start with your workflows and your data, not with a tool they want to sell you. Some only advise. Others also build and run what they recommend. Knowing which you need is the first decision.

What an AI consultant actually does

The work usually covers some or all of these:

  1. Finding the opportunities. Mapping how work happens today and spotting the tasks that are frequent, rule-based and digital enough for AI to help.
  2. Checking feasibility. Testing whether AI can do the task well enough, using your own real examples, before anyone commits to a build.
  3. Choosing the approach. Off-the-shelf AI features, an automation tool, or AI built into your systems.
  4. Data and risk. Where your data goes, which providers are used, what needs a person to approve, and what your obligations are.
  5. Building it, if the consultant also delivers, or briefing the people who will.
  6. Training the team so the new way of working sticks.
  7. Looking after it, because AI features need re-testing when models, prices or your process change.

The three kinds of AI consulting

Strategy-only consultants produce recommendations and a plan. Useful for large organisations with their own engineering teams. For a small business, a plan without anyone to build it often stalls.

Tool resellers and implementers set up a specific product. Quick if that product fits your workflow. Less useful if it doesn't, because the recommendation starts from the tool.

Build-and-run studios scope the workflow, build AI into your systems, test it, and look after it once it is live. They suit businesses that want an outcome rather than a report.

Do you need an AI consultant?

You probably do if:

  • Staff use AI tools on their own, but nothing is connected to your systems.
  • You've tried a chatbot or AI tool and it stalled.
  • You're not sure what company data is safe to use with AI.
  • You know roughly what you want, but not whether it is realistic or what it would take.

You probably don't if a tool you already pay for has AI features that cover your need. Try those first.

Questions to ask before you hire one

  1. Do you start from our workflows or from a product? You want the former.
  2. How will you test whether the AI is good enough? Look for a clear answer involving your real examples.
  3. What happens when the AI is wrong? There should be a plan for review and correction.
  4. Where will our data go? Ask which providers are used and whether they train on your data.
  5. Who builds it, and who looks after it afterwards? A plan with no owner after launch is a risk.
  6. Who owns what you build? You should own the code and your data.
  7. Can we see work you've done? And is it clearly labelled: client work, or their own products?
  8. How do you price? Be wary of a fixed price given before anyone understands your needs.

How we work

We are a build-and-run studio. We don't give a price until we understand what you need: it starts with a free consultation about how your work happens today. For complex projects we suggest a short paid Scoping Sprint first, with a clickable prototype and a fixed price, credited to the build. Then we build in stages you can see, test any AI on your real examples, and look after it once it is live. You own the code.

If you'd like to see how we think about a first AI project, take the AI Workflow Check, or book a free consultation.

What a good first engagement looks like

A sensible first AI project for a small business follows roughly this shape:

  1. A conversation about how work happens today. Who does what, in which tool, and where it breaks.
  2. One workflow chosen, the one that is frequent, rule-based and costly enough to be worth it.
  3. A feasibility test on your real examples, so everyone knows whether the AI is good enough before a build starts.
  4. A clear scope and a fixed price for the first version, including who reviews what and how it will be measured.
  5. A build in stages you can see, with the AI tested against the agreed bar before launch.
  6. Launch, training and a period of close support, then ongoing care: monitoring, re-testing and model updates.

Each step reduces risk before the next one costs more.

Red flags

  • A recommendation before anyone has looked at your workflows.
  • No plan for testing the AI or for what happens when it's wrong.
  • Vague answers about where your data goes.
  • A big programme proposed before a single workflow has been proven.
  • Client names and results you can't verify, or in-house products presented as client work.

Frequently asked questions

What's the difference between an AI consultant and an AI developer? A consultant decides what should be built and why. A developer builds it. Some firms do both, which avoids a hand-over gap.

How long does an AI consulting engagement take? It depends on the scope. A focused first workflow is much quicker than a company-wide programme, which is one reason to start with one.

Do we need our own technical team? Not if the firm builds and looks after what it recommends. If they only advise, you will need someone to build it.