What Questions Should I Ask Before Hiring an AI Consultant?

31 March 2026

1. "How will we measure whether this project succeeded?"

This is the single most important question, and the one most businesses forget to ask until it is too late.

A good consultant will define success in terms you already care about: cost reduction in pounds, time saved in hours, revenue generated, error rates reduced, customer satisfaction scores improved. They should be specific. "We will reduce your invoice processing time from 12 minutes to under 2 minutes per document" is a good answer. "We will leverage AI to optimise your workflows" is not.

Watch out for consultants who define success in technical metrics that mean nothing to your business. Model accuracy percentages, F1 scores, and benchmark results are important for engineers but useless as standalone success criteria. If they cannot translate technical metrics into business outcomes, they either do not understand your business or they are hiding behind jargon.

Red flag: "Success will depend on many factors and is difficult to guarantee." Every project has uncertainty, but a consultant who cannot commit to measurable outcomes is one who cannot be held accountable.

2. "What happens to our data during and after the project?"

Your data is the most valuable thing in the engagement. Be very specific about:

During the project: Where is your data stored? Is it encrypted at rest and in transit? Who has access? Is it processed within UK jurisdiction? Will it be used to train any models that other clients might benefit from?

After the project: Is your data deleted? How can you verify deletion? If they fine-tuned a model on your data, who owns that model? Can they use the patterns learned from your data in other engagements?

A trustworthy consultant will have clear data handling policies in writing before you ask. They will not be offended by these questions. They will welcome them because they demonstrate that you take data governance seriously.

Red flag: Vague answers about data handling, or policies that are "still being finalised." If they have not figured out data governance by now, they are not ready to handle your data.

3. "Who owns the intellectual property you create for us?"

This catches more businesses than you would expect. Some AI consultants retain ownership of custom models, code, and configurations they build for you. Others grant you a licence to use them but retain the underlying IP. The best ones assign full ownership to you as part of the contract.

Be especially careful with fine-tuned models. If a consultant fine-tunes a model on your proprietary data and retains ownership of the resulting model, they effectively own a version of your knowledge.

Get this in writing before the project starts. Not in a verbal agreement, not in an email. In the contract.

Red flag: "We retain ownership of our proprietary methodologies and any models developed during the engagement." This means you are paying to build their asset, not yours.

4. "Tell me about a project that failed. What went wrong?"

Every experienced AI consultant has had projects that did not work. AI is probabilistic, data is messy, and business requirements change. If a consultant tells you they have never had a project fail, they are either lying or they have not done enough work to be credible.

What you are listening for is honesty, self-awareness, and learning. Did they identify why it failed? Did they flag the risk early? Did the client get a fair outcome? How did they change their approach as a result?

A consultant who can discuss failure openly is one who will be honest with you when your project hits obstacles, and every project hits obstacles.

Red flag: "All our projects have been successful." No, they have not. Next.

5. "How do you price your work, and what is not included?"

AI consulting pricing in the UK varies enormously. Day rates for experienced consultants range from £800 to £2,500. Project-based pricing for a typical SMB engagement might be £15,000 to £75,000 depending on scope and complexity.

More important than the headline price is what is included and what is not. Common hidden costs include:

Infrastructure. Does the quoted price include cloud compute costs, or are those billed separately? For AI workloads, compute can be substantial.

Data preparation. Cleaning, formatting, and labelling your data is often the most time-consuming part of an AI project. Some consultants include this; others treat it as a separate billable phase.

Ongoing maintenance. AI models degrade over time as the data they were trained on becomes stale. Who maintains the model after deployment? At what cost?

Change requests. What happens when requirements evolve mid-project (they always do)? Is there a change request process, and how is it priced?

Ask for a detailed breakdown, not just a total. And ask what happens if the project runs over budget. A consultant who has done this before will have clear answers.

Red flag: "We will scope that once we get started." This means the price will increase once you are committed.

6. "What happens after the project ends?"

This question reveals whether the consultant is building something sustainable or creating a dependency.

A good consultant builds solutions your team can maintain, or at minimum, clearly documents what ongoing maintenance is needed and what it will cost. They should offer knowledge transfer, documentation, and training as standard parts of the engagement.

A consultant who builds a black box that only they can maintain has a financial incentive to keep you dependent on them. That is not inherently wrong, managed services are a legitimate model, but you should know what you are signing up for.

Ask specifically: can we bring maintenance in-house after the project? What skills would our team need? Will you provide documentation sufficient for another consultant to take over if we choose?

Red flag: "Our solution requires our proprietary platform for ongoing operation." This is vendor lock-in by another name.

7. "Is there a scenario where you would tell us NOT to use AI?"

This is the honesty test. Not every business problem needs AI. Sometimes a spreadsheet, a better process, or a simple automation tool is the right answer. A consultant who only sells AI will recommend AI for everything, even when it is the wrong solution.

The best consultants will tell you when AI is not appropriate. They will say things like: "For your current data volume, a rules-based system would be cheaper and more reliable." Or: "You need to fix your data collection process before AI can add value here."

A consultant who is willing to talk themselves out of a sale is one you can trust with the projects where AI genuinely is the right answer.

Red flag: "AI can solve virtually any business challenge." No, it cannot. And anyone who says otherwise is prioritising their revenue over your results.

Is This Right for You?

If you are a business owner or leader evaluating AI consultants for the first time, these seven questions give you a solid framework. You do not need to be technical. You need to be thorough.

If you are already working with a consultant and have not asked these questions, it is not too late. A good consultant will be glad you asked. A bad one will be uncomfortable, which tells you what you need to know.

If your organisation has strong internal technical capability and a clear problem definition, you may not need a consultant at all. Many AI implementations can be handled by upskilling existing teams, especially for well-documented use cases with available open-weight models.

A Final Note on Bias

We are an AI consultancy writing about how to hire AI consultants. We have an obvious interest here. Take our advice, weigh it against other sources, and use your own judgement. The fact that we are telling you to be sceptical, including of us, is the most honest thing we can say.

Frequently Asked Questions

How much does AI consulting cost in the UK?

Day rates for experienced AI consultants in the UK range from £800 to £2,500. Project-based engagements for SMBs typically cost £15,000 to £75,000 depending on scope. Always ask for a detailed breakdown including infrastructure, data preparation, and ongoing maintenance costs.

How do I know if an AI consultant is legitimate?

Ask about their failures, not just successes. Request references from previous clients in your industry. Check whether they can explain their approach in plain language without hiding behind jargon. A legitimate consultant will welcome scrutiny and have clear answers about data handling, IP ownership, and pricing.

Should I hire an AI consultant or build an in-house team?

For your first AI project, a consultant can accelerate delivery and reduce risk. For ongoing AI work, building internal capability is usually more cost-effective long-term. Many businesses start with a consultant for the initial implementation and use the engagement to upskill their own team for future projects.

What are the biggest red flags when hiring an AI consultant?

Vague success metrics, unclear data handling policies, retained IP ownership of work you paid for, inability to discuss past failures, pricing that will be scoped after you commit, solutions that only they can maintain, and claims that AI can solve any problem. Any of these should prompt further scrutiny.