How to choose an artificial intelligence consultancy: 10 questions to ask before hiring

Every week a new company pops up promising to bring artificial intelligence to your business. Some of them will do it well. Others will leave you with a spectacular demo that never makes it into day-to-day use. The difference almost never lies in the technology they claim to master, but in how they work before, during, and after the project.

This guide brings together the questions worth asking before you sign anything, the red flags worth taking seriously, and what a good proposal should include. You can use them to compare any provider, including us.

Before you start looking: be clear about what you want to solve

Your first conversation with a consultancy will be far more useful if you come with a specific problem rather than a general wish. “We want to implement AI” is not a goal. “Sort the customer requests that come in by email, which currently take up several hours a day” is.

You don't need to have the solution. It's enough to have identified two or three processes that cost you time, errors, or customers, and to have a rough idea of what they cost today. With that, a good consultancy can already tell you whether AI is the right tool or whether there's a simpler solution.

10 questions to evaluate an AI consultancy

1. Do they start with your processes or with the tool?

If they're already showing you a specific platform in the first meeting, be wary. The right order is to understand your operations, find where the real cost lies and, only then, choose the technology. A consultancy that always proposes the same tool isn't diagnosing anything: it's selling.

2. Can they show you projects running in production?

A demo proves that something is possible. A project in production proves that it's being used, that it holds up under real-world volume, and that someone maintains it. Ask for examples of systems that have been running for months, and ask what problems came up after go-live and how they were solved.

3. Who will work on your project?

Ask about the specific team: who will do the assessment, who will handle development, and who will provide support. It's worth knowing whether the work is done by an in-house team or subcontracted, and whether there are business profiles as well as technical ones. An AI project fails more often because the process isn't understood than because of a lack of coding skills.

4. How will it integrate with the systems you already use?

AI adds value when it works with your CRM, your ERP, your email, or your documents, not in an isolated application. Ask how it will connect to what you already have, whether data will need to be migrated, and what happens if you replace one of those tools tomorrow.

5. Where will your data be, and how is GDPR compliance ensured?

This is one of the most important questions, and one of the least asked. Ask them to explain where the data is processed, which providers are involved, whether the model provider can train on your information, and what data processing agreements are signed. If the answer is vague, don't go any further.

6. How will success be measured?

Before starting, both of you should agree on which business metric needs to improve: hours saved, response time, errors, recovered sales. Without a baseline measurement, there's no way to know whether the project has worked, and the conversation ends up relying on impressions.

7. What happens after go-live?

An AI system needs ongoing follow-up: reviewing responses, adjusting instructions, updating documents, and keeping an eye on costs. Ask what support the proposal includes, how the system is monitored, and who is responsible for improving it based on real-world use.

8. Are they independent of model providers?

OpenAI, Anthropic, Google, and Meta release new models all the time, and each one is a better fit for some use cases than for others. An independent consultancy chooses the model based on your use case and designs the solution so the model can be swapped out if a better or cheaper option comes along tomorrow.

9. How do they manage risks and EU regulations?

The EU Artificial Intelligence Act is already in force and, among other things, requires that people be informed when they are interacting with an AI system. Beyond regulations, ask what limits the system will have, which actions will require validation by a person, and how everything it does will be logged.

10. Can they help you with funding?

Public digitalization grants open and close their calls for applications, so it's not wise to plan a project counting on a subsidy that doesn't exist today. Even so, it's useful for the consultancy to know the available funding programs and let you know when one opens. You can check the current status of digitalization grants before deciding.

Red flags

• They promise you a specific return before analyzing your processes.
• They propose “AI for everything,” even for tasks that simple automation would handle better.
• They don't ask about your data, who uses it, or where it's stored.
• The quote doesn't detail the scope, the deliverables, or what happens after go-live.
• The solution only works with their platform, and it isn't clear how you would get your data back if you decide to switch.

Does it matter whether the consultancy is nearby?

A lot of the work can be done remotely, but proximity helps in the phases that have the greatest impact on the outcome: the sessions to understand your processes, the testing with the teams that will use the system, and the training. It also makes it easier for the consultancy to know the local business landscape and the support programs in your area. What matters is that the format adapts to you: in person, remote, or a combination of both.

How we work at Namastech

We are a technology consultancy based in Barcelona, and since 2006 we have worked with more than 480 companies and institutions. We always start with a no-obligation assessment to identify which processes would benefit from AI and which ones don't need it. We then design the solution, integrate it with your systems, train your team, and measure the results so we can keep improving it.

If you'd like to learn more, you can see how we approach technology and AI consulting or why most AI projects never make it to production.

Frequently asked questions

How much does it cost to hire an artificial intelligence consultancy?

It depends on the scope: an assessment doesn't cost the same as developing an AI agent and integrating it with several systems. The most prudent approach is to start with a focused assessment that identifies the process with the best effort-to-impact ratio, and then ask for a quote with deliverables and measurable results.

How long does it take to see results?

If the project is well scoped, weeks rather than months. Projects that take a long time to deliver results are usually the ones that started with an overly broad goal.

Do I need to have my data perfectly organized before getting started?

No. The assessment is precisely what shows which data is needed and what state it's in. The usual approach is to start with a process where the information already exists and to improve data quality as the project progresses.

What is the difference between an AI consultancy and a software vendor?

A software vendor sells its own product. A consultancy analyzes your case, recommends the right solution, which may be a third-party product, custom development, or no AI at all, and takes care of integrating it and making it work with your processes.