How to Choose an AI Consultant: 7 Questions to Ask

How to Choose an AI Consultant

Choosing an AI consultant is harder than it looks. The market is full of people who sell strategy, charge well for it, and leave you with a document you cannot implement. Knowing how to choose an AI consultant correctly means asking the right questions before you sign, not after you have spent $30,000 on a slide deck.

This guide gives you seven concrete questions to run through with any consultant you are evaluating, the red flags that signal a bad fit, and a clear picture of what good deliverables actually look like.

Why most AI consultants cannot build what they sell

The AI consulting market split into two camps early. The first camp is strategy consultants who understand AI conceptually. They can map your processes, identify opportunities, and produce a roadmap. The second camp is implementers who actually build systems, encode your IP, and ship something that runs.

Most consultants operate in the first camp and present themselves as both.

The gap matters because a roadmap without implementation is not a business outcome. For $1M-plus knowledge businesses, the specific problem is that your delivery model depends on expertise that lives in your head. A generic AI strategy does not address that. A system built around your actual methods does.

The AI consulting services that produce real results are the ones scoped around your specific IP, not a standard set of tools applied to a generic workflow. Before you evaluate anyone, decide what you actually need: a plan, or a working system.

Independent research backs this up. The Stanford AI Index shows AI adoption rising fast while measurable business value lags, precisely because deployment and integration are where most efforts stall. A consultant who cannot get past the strategy stage will not close that gap for you.

Seven questions to ask before hiring an AI consultant

Run through these before any engagement conversation gets to scope or pricing. If a consultant cannot answer them clearly, keep looking.

1. Can you show me a working deliverable from a similar client?

Not a case study. Not a testimonial. An actual system, automation, or prototype that runs. Ask what it does, how it was built, and whether the client is still using it. The answer tells you whether they build or whether they advise.

2. What does your engagement scope include, and what is explicitly excluded?

Good consultants are precise about scope. They can tell you what they will build, what they will not build, and why the boundary sits where it does. Vague answers here mean vague deliverables later.

3. What milestones will I see in the first 90 days?

A working prototype is achievable in 90 days for most knowledge business use cases. If the first milestone is a completed discovery phase at week twelve, the engagement is structured around billable time rather than outcomes. Push for specifics: what will exist, what will it do, and when.

4. How do you encode my specific IP, not just apply general AI tools?

This is the question that separates orchestration from automation. General AI tools applied to a generic process produce generic output. A system built on your methodology, your language, and your decision logic produces output that reflects your expertise. Ask how they plan to capture that before they build anything. Strong builders work to the same principles Anthropic sets out for building effective agents: clear scope, defined inputs, and explicit quality criteria, not open-ended prompts.

5. What is your implementation model: done-for-you or done-with-you?

The answer shapes everything about how the engagement runs and what you get at the end. More on this in a later section. For now, make sure you understand which model they use and why they recommend it for your situation.

6. Who builds the work, and will I have access to them?

Some consultants sell the engagement and hand delivery to a junior team or a subcontractor you have never met. If the person you are talking to is not the person building your system, find out who is and ask to meet them before signing.

7. What does success look like, and how will you measure it?

If they cannot name a specific outcome, a specific metric, and a timeframe for measuring it, the engagement has no accountability structure. A clear answer might look like: "In 90 days you will have a working intake and onboarding system that reduces your manual time by a set number of hours per client." Vague answers mean no commitment.

Red flags to watch for

Some of these are obvious in hindsight. They are harder to see when someone is confident and articulate in a sales conversation.

• Tool-first framing. If the pitch centres on which AI tools they use rather than what your business needs to achieve, they are selling a product, not solving your problem.
• No milestone structure. Month-to-month engagements without defined deliverables are advisory retainers dressed as implementation work.
• Reluctance to show real work. A consultant who cannot produce evidence of past deliverables either has not built much or is under NDA for everything. One NDA is plausible. A portfolio of zero visible work is not.
• Generic case studies. "We helped a consulting firm increase efficiency" tells you nothing. Ask for specifics: what was built, what changed, what the outcome was in measurable terms.
• Overpromising on speed. AI systems built on your IP take time to build properly. Anyone promising a complete transformation in four weeks is compressing the work in ways that will show up as quality problems later.

You can find a detailed breakdown of what to look for across different provider types in our AI consulting firms comparison.

What good deliverables look like

The deliverable at the end of a genuine AI consulting engagement is not a document. It is a system that runs.

For a knowledge business, that typically means one or more of the following:

Deliverable	What it does
IP extraction document	Your methodology captured in structured, trainable format
Working AI agent or workflow	Automates a specific part of your delivery model
Scaling roadmap with build specs	Not just what to do, but how it will be built and by whom
Testing and validation results	Evidence the system performs against your quality standard
Handoff documentation	So your team can operate and iterate without the consultant present

A strategy deck alone is not a deliverable. A list of tools is not a deliverable. If the engagement scope does not include at least one thing that runs independently of the consultant, the value is advisory only.

For more on what different engagement types cost and what you get at each level, the AI consulting cost and pricing guide breaks this down in detail.

Done-for-you vs done-with-you: which fits your business

This is the question most buyers overlook, and it is often more important than which consultant you choose.

Done-for-you vs done-with-you is not just a delivery format. It determines how much of your IP gets encoded into the system and how well your team can operate it after the consultant leaves.

Done-for-you works when the problem is well-defined and your team will not need to adapt the system much. The consultant builds it, you receive it. Faster to scope, easier to price.

Done-with-you works better for knowledge businesses where the output depends on your specific methodology. The consultant works alongside you over weeks, capturing how you think, how you make decisions, and how your delivery actually runs. The system that comes out of that process is yours in a way that a built-and-handed-over system is not.

For most $1M-plus educators and consultants, done-with-you produces better long-term results. The IP encoding is deeper. The team understands how the system works because they watched it being built. And the consultant can course-correct in real time as they learn more about your actual model.

The trade-off is that it requires more of your time during the engagement. Plan for two to four hours per week of active involvement, not passive review.

What to do before you hire

Before any consultant engagement starts, get three things in order.

Map your delivery model. Write down, as specifically as you can, how a client moves from first contact to final outcome. Every step. Every decision point. Every place where your personal judgment is currently required. This is the raw material the consultant will work with. If you cannot articulate it, the consultant will spend expensive time figuring it out.

Define one clear problem to solve first. Not ten things. One. The best AI consulting engagements start narrow and expand. Pick the part of your delivery model that is most repetitive, most time-consuming, or most dependent on you personally. That is where the first build should focus.

Set a clear success metric. What would make this engagement clearly worth it? Put a number on it. Time saved per week, clients served per month, hours of founder involvement per engagement. If you cannot define success before you start, you cannot evaluate it when the engagement ends.

Start with the assessment

If you are not sure where AI would have the most impact on your delivery model, that is exactly what the IP Monetisation Assessment is designed to answer. It maps your current delivery against the three phases of the Explore, Map, Transform method and identifies where the highest-value build opportunity sits.

Take the IP Monetisation Assessment before you speak to any consultant, including us. You will get a clearer picture of your starting position, and any qualified consultant should be able to work with the output directly.