How to Choose an AI Consulting Company: 9 Criteria
Short answer
Choose an AI consulting company by what they've delivered, how they start, who does the work, how they handle your data and what you own at the end. Be wary of anyone who leads with tools, promises big returns or can't name a real result.
I run a business that does AI consulting work, so you should know my bias up front. This isn't a list of who to hire, and it doesn't put me first. If you're about to hire an AI consulting company, it's the checklist I'd use if I were buying.
What types of AI consulting companies are there?
Five types: large consultancies, boutique firms, fractional leaders, build agencies and training providers. They solve different problems.
| Type | Good for | Watch for |
|---|---|---|
| Large consultancy | Big programs, complex governance | Cost, junior staff doing the work |
| Boutique AI firm | Focused projects, senior attention | Narrow range of skills |
| Fractional leader | Ongoing direction without a full-time hire | You still need people to do the work |
| Build agency | Custom software and agents | Building before the problem is clear |
| Training provider | Getting your team capable | No delivery after the course |
Know which one you need. If you want a custom AI agent built, you want a build agency. If you want direction, you want a leader. If you want your people trained, you want a trainer. I explain the leader versus project difference in fractional chief AI officer vs AI consultant.
What are the nine criteria?
Proof, start point, who does the work, data handling, ownership, measures, honesty, fit and exit.
1. Proof of real work. Ask for one example of AI work that ran in a real business. No names needed. What was the problem, what changed and how did they know?
2. Where they start. The good ones start with your problem, not their tool. If the first meeting is a demo, be careful. I explain why in business-first AI.
3. Who does the work. The person who sells you is often not the person who delivers. Ask who will do the work and meet them.
4. How they handle your data. Ask what data they'll use, where it goes and what the AI tools do with it. Australia's Guidance for AI adoption gives a plain checklist, including accountability and human control. NIST's AI Risk Management Framework, published 26 January 2023, goes deeper on managing AI risk.
5. What you own at the end. Do you own the instructions, the workflows and the data? Or are you renting them? Get this in writing.
6. How they measure success. Good consultants agree on a number before they start. Time saved, cost cut, revenue added.
7. How honest they are. Ask what AI can't do for you. If they have no answer, walk.
8. Fit with your size. A firm built for large companies may not suit a business of 20. The reverse is also true.
9. How you can leave. Is there a clean end point? Avoid open-ended retainers with no outcomes.
The questions I'd want a buyer to ask me
These are questions I've heard from business owners. I'd rather you asked them early.
- How do I use AI without risking my reputation or client confidentiality? If a firm brushes this off, walk.
- How do I know it isn't another shiny tool that won't work for me? Ask them to start from your problem, not their tool.
- I'm already drowning. How do I add AI without adding to my plate? Ask what they'll take off your desk, and who does it.
- Will this work for my type of business? Ask for one example close to yours.
One thing changed how I work. When I first built my offer, I led with tools and automations. People got excited, then confused. Every conversation hit the same line: "but my business is different." So I flipped it. I now teach the method first, and show the tools after. Once owners understood the method, that objection fell away. I look for the same thing in any firm I'd hire. They should explain how they think before they show you what they sell.
How do you build a shortlist in a week?
Write the problem, find three firms of the right type, ask each the same questions, then score them on the nine criteria.
- Day one: write the problem. One paragraph. What's slow, costly or stuck, and what number would prove it's fixed.
- Day two: pick the type. Use the table above. A direction problem and a build problem need different firms.
- Day three: find three firms. Ask people you trust, look at who writes plainly about real work and check how each firm describes its own results.
- Days four and five: take the calls. Ask each firm the same questions, listed below. Take notes.
- Day six: score them. Give each firm a score from one to five on each of the nine criteria. Don't weigh by who you liked most.
- Day seven: decide. Pick the top one or two and ask for a written scope.
Here's an illustrative example, not a real search. Say you run a 20-person services firm. Firm A has a great demo but can't say who does the work. Firm B is smaller, names the person who'd deliver and agrees to one measure in writing. On the nine criteria, B wins on who does the work, measures and honesty. A wins only on polish. Polish isn't on the list for a reason.
What are the red flags?
Leading with tools, guaranteed returns, no named result and pressure to sign fast.
- "We'll transform your business with AI." No measure, no start point.
- Guaranteed ROI. Nobody can guarantee your outcome.
- No examples. Or examples with no numbers.
- Jargon. If they can't explain it plainly, they may not understand it.
- Pressure. A good firm gives you time.
If the company's pitch is about tools and not your problem, that's the same pattern behind why AI implementation fails.
What questions should you ask in the first call?
- What problem would you start with in my business, and why?
- Who will do the work?
- What's a similar project you've delivered, and what changed?
- What data would you need, and how do you protect it?
- What will I own at the end?
- How will we measure success?
- What could go wrong?
- What happens after the first project?
If you also want a plain description of the job, read what does an AI consultant actually do.
What should a good proposal include?
The problem in your words, the first project, who delivers it, one measure, what you own, the price and the exit.
If a proposal is missing any of these, ask for it before you sign:
- The problem, in plain words. If they can't restate your problem, they didn't listen.
- A first project with an end date. Not "ongoing support".
- Named people. Who does the work, and how much of their time.
- One agreed measure. The number that says it worked.
- Ownership. What's yours at the end: instructions, workflows and data.
- Data terms. Where your data goes and who can see it.
- A clear price and exit. What it costs, what's extra and how either side can stop.
How should you compare quotes?
Compare outcomes, not hours. Two quotes with different scopes can't be compared on price. Ask each firm to write the one measure they'll hit. Then compare the measure, the owner and the exit.
Where do you look if you want my help?
I'm open about what I do. Njin builds AI systems for Australian businesses, and I run AI consulting and training through James Killick Co. If you'd rather have a team build custom software or agents for you, that's Devwiz. Judge us with the same nine criteria.
FAQ
How much does an AI consulting company cost?
It varies a lot with scope and type of firm, so I won't give a number. Ask each firm for a fixed scope, one measure and a price for the first project. Then compare those, not hourly rates.
Do I need an AI consultant or an AI agency?
If you need direction, a plan or training, you want a consultant or a trainer. If you want custom software or an agent built, you want a build agency. Many problems need direction first, then a build.
How long does an AI consulting project take?
It depends on the scope. Ask for the first project to have a clear end date and a measure. Be wary of open-ended work with no outcome.
Should I pick a big firm or a small one?
Match the firm to your size and your problem. Big firms suit complex governance. Small firms suit focused projects with senior attention. Use criterion eight to check fit.
What if I'm not sure what to ask for?
Start with the problem and the number. If you can't write those down, a good firm will help you find them in the first call.
Your next step
Before your first call, write down the one problem you'd want a consultant to solve and the number that would prove it worked.
About the Author
The AI Orchestrator
AI Orchestrator and entrepreneur with 10+ years building digital products. Helping $1M+ business owners scale with AI systems, automation, and implementation through The AI Orchestrators, Devwiz, and Njin.