What questions should I ask an AI consultant before hiring?

It depends on what you need them for. If you mainly need training, ask how they teach, whether they use your real work and how they measure adoption. If you need strategy, ask how they find the right use cases and what you will have at the end. If you need something built, ask who writes the code, how your data is handled and who maintains the system afterward. For any of these, ask who will actually do the work and how they define success before you sign.

Who this is for

CEOs, COOs, managing partners and operations leads who are speaking with AI consultants or firms and want to compare them on substance rather than on the quality of their pitch.

The short version

Ask about the work itself: who does it, how, and what you own at the end.
Ask how success will be measured, in terms you care about.
Watch for vague answers, tool-first pitches and no plan for after they leave.

Before you talk to anyone

Write down the problem you want solved in a few sentences. “We want to use AI” is not enough. “Our team spends too long preparing client reports” or “nobody uses the AI licenses we bought” gives a consultant something concrete to respond to, and makes their answers easier to compare. If you are unsure what kind of help you need, our guide to training vs consulting vs engineering is a good place to start.

Questions about their approach

How would you approach our problem in the first few weeks? A good answer is specific to your situation and starts with understanding your work, not with a product.
How do you decide which tasks are worth using AI for? Look for a clear method that weighs time spent, risk and how easy the result is to check. See where to start with AI.
What would you advise us not to do? Honest consultants will tell you where AI is a poor fit. If everything is a good fit, be cautious.
Do you recommend specific tools, and do you have any commercial relationship with them? Partnerships are not bad in themselves, but you should know about them.

Questions about who does the work

Who will actually be working with us? At some firms, senior people sell the work and junior people deliver it. Ask to meet the people who will be in the room.
Have they done this kind of work before? Ask them to describe similar engagements in plain terms. Reasonable firms protect client names, but they should be able to explain what they did and what changed.
Who can we speak to? References from past clients are a fair request for any significant engagement.

Questions about data and security

Which tools would our data go into, and on what terms? You want business plans with clear data protections, not personal accounts.
Will you need access to our systems, and how is that controlled? Ask how access is granted, limited and removed at the end.
How do you handle confidential client information? The answer should match or strengthen your own rules. Our page on keeping client data safe with AI covers what good practice looks like.

Questions about what you get

What exactly will we have when you finish? Trained people, written policies, a roadmap, working software, documentation, or some mix. Get it in writing.
Who owns what you build? Code, prompts, configurations and documents created for you should normally belong to you. Check the contract.
What happens after you leave? Ask who maintains any systems, how your team is prepared to own them and what ongoing support costs. See who maintains AI systems after the consultant leaves.
How do you build capability inside our team, rather than dependence on you? Good partners plan to make themselves less necessary over time.

Questions about success and cost

How will we know this worked? Agree on measures before the work starts, such as time saved on a named task, usage of approved tools or a working system in daily use.
How is the work priced, and what could change the price? Fixed scope, time-based and retainer models each have trade-offs. What matters is that you understand what is included.
Can we start small? A short first phase, such as a training session or a pilot, lets you judge the fit before a larger commitment. See how to run an AI pilot that doesn’t stall.

Warning signs

The pitch starts with a tool, not your problem. Licenses alone rarely change how a firm works; see why licenses are not a strategy.
Big promises with no method. Confident numbers about savings, without explaining how they will be measured.
No mention of training or change management. Systems people do not use do not create value.
Vague answers on data. If they cannot say clearly where your data goes, keep looking.
No plan for the handover. If everything depends on them indefinitely, that is a cost to plan for.

Common mistakes

Comparing on price alone. A cheaper engagement that nobody adopts costs more in the end.
Skipping the scope. Without a written scope and measures, it is hard to judge results.
Hiring for engineering when the problem is adoption. Many firms need their team trained before anything is built.
Not involving the people who will use the work. Bring in a team lead who does the job day to day.

When you do not need us

If your need is narrow, such as setting up one approved tool with good data rules, and you have someone internally who can lead it, you may not need a consultant at all. Our guide on hiring an AI lead vs a consultant can help you decide.

How Ruby helps

AI training: hands-on training built around your team’s real work, with clear measures of adoption.
AI strategy: finding where AI belongs in your firm, setting policies and running measured pilots.
AI engineering: building custom agents and integrations that you own, with a clear plan for maintenance.

We are happy to answer every question on this page. To ask them directly, book a call with Ruby.