Choosing an AI consultancy is hard because almost every firm now claims AI expertise. The questions below help you separate partners who will genuinely help you change from those who will deliver a polished deck and leave. For each, we have included what a good answer sounds like and the red flags to listen for.
About their experience
1. What AI work have you delivered for organisations like ours? Good: specific examples in your sector or size, with outcomes. Red flag: only big-name logos, no detail.
2. Can we speak to two recent clients? Good: yes, readily. Red flag: hesitation or only very old references.
3. Which of your projects did not work, and why? Good: an honest example and what they learned. Red flag: "all our projects succeed".
About their approach
4. How will you decide which use cases we should prioritise? Good: talking to our teams, looking at real workflows, assessing value and feasibility. Red flag: a standard list they apply to everyone.
5. Do you recommend starting small? Good: yes, with pilots and clear success criteria. Red flag: pushing a large programme from day one.
6. Are you tied to any AI vendors? Good: transparent disclosure of partnerships. Red flag: always recommending the same platform.
7. How do you handle data protection and UK GDPR? Good: a clear process, involving our data protection lead. Red flag: vague reassurance.
About delivery and handover
8. Who exactly will do the work? Good: named people you meet before signing. Red flag: senior partners pitch, juniors deliver.
9. What will we have at the end that we did not have before? Good: specific deliverables and capabilities. Red flag: "a strategy" with no detail.
10. How will you transfer knowledge to our team? Good: a plan for training and documentation. Red flag: no plan, which means dependence.
11. Who leads change management and staff adoption? Good: they support it, but expect an internal leader to own it. Red flag: ignoring adoption entirely.
About measurement and cost
12. How will we measure whether this worked? Good: baselines, targets and review points agreed upfront. Red flag: "AI benefits are hard to measure".
13. What will this cost in total, including follow-on work? Good: clear pricing and honest estimates of later phases. Red flag: low entry price, vague later costs.
14. What could we do ourselves instead? Good: honest advice about which parts you could handle internally. Red flag: everything requires them.
15. What would make you recommend we do not proceed? Good: thoughtful conditions, such as no clear business problem. Red flag: nothing would.
Before you even start interviewing
The best protection is having someone internal who understands AI well enough to judge the answers. Without that, it is hard to tell a strong proposal from a generic one. Our Leading AI Innovation & Organisational Transformation course builds exactly that understanding in a non-technical manager.
For the broader selection process, see our guide on choosing an AI consulting firm, and for the scope question, AI consulting vs AI implementation.
Frequently asked questions
How many AI consultancies should we interview?
Three is usually enough to compare approaches and prices without the process dragging on.
Should we pay for a small paid discovery phase first?
It can be a good way to test working together before committing to a larger contract, as long as the deliverables are clear.
What is the biggest mistake when hiring an AI consultancy?
Hiring without an internal owner. Even the best consultants cannot make your organisation change if nobody inside is accountable for it.