The title is unregulated, which means it covers people doing genuinely different jobs. Knowing which one you are hiring is most of the value of the decision.
The three things sold as "AI consulting"
Advisory. Someone reviews how the business actually works, finds where hours and errors go, and tells you what to do about it. The deliverable is a prioritised list and the reasoning behind it. No software is built.
Enablement. Training your team on tools they will use, with prompts and playbooks they keep. The deliverable is capability inside your business.
Delivery dressed as advice. A discovery phase whose real function is to justify a build. Not automatically wrong, but you should know that is what it is, because the recommendation is rarely going to be "do nothing".
What a good one actually does
Spends most of the engagement on your business rather than on AI. The hard part is never the technology — it is working out which of your processes are stable enough to be worth encoding, which are about to change anyway, and which are broken in a way software would only entrench.
They should also be able to tell you what not to automate, and to say when a problem is a management problem wearing a technology costume. A consultant for whom everything is an AI opportunity is selling capacity rather than judgement.
When you do not need one
If you already know which process is costing you and roughly what a fix looks like, you need a build partner, not an advisor. Paying for a discovery phase to confirm what you already know is a common and avoidable expense.
You also do not need one if the honest answer is that nothing is stable enough to automate yet. A business mid-restructure is better served fixing the process than encoding it.
How to judge one before you commit
Ask what they would tell you not to do, and ask what would make them recommend walking away. Specific answers come from people who have run the engagement before. Uniformly positive answers usually mean the recommendation was decided before the review began.
Questions
- What is the difference between an AI consultant and an AI agency?
- A consultant sells judgement and hands you a decision; an agency sells delivery and hands you a working system. Many firms do both, which is fine as long as the advisory stage can honestly end in "do not build this".
- How long should an engagement be?
- An initial audit is usually a matter of weeks, not months — long enough to see how the work really happens, short enough that the findings are still current when they land. Ongoing advisory works better as a small standing commitment than a large one-off.
- Should the consultant also build the system?
- It can be efficient, because the context does not have to be rebuilt by a second team. The thing to watch is whether the advice can still be honest when the same firm profits from a yes. Ask that question out loud; the answer is revealing.
Wondering whether this applies to your business? That is a thirty-minute call, not a research project.
Book a Consultation Call