There is a pattern that repeats itself frequently in mid-size companies that engage AI services: three months in, the project is "in progress," the internal team doesn't know how to operate what was built, and the consulting firm keeps billing to maintain something nobody fully understands.
This isn't bad faith. It's a lack of clarity from the start.
The questions you ask before signing largely determine what happens afterward. This checklist is designed for COOs who need to make an informed decision — not for those looking to validate one they've already made.
What concrete deliverables are committed within the first six weeks?
A serious proposal must specify what will be running in production — not in a demo — within the first few weeks. If the answer is "it depends on the discovery process," ask them to spell it out in the contract.
What you should see: at least one operational agent or workflow, documentation of the automated process, and evidence that the internal team participated in the build.
What should concern you: proposals that speak of "phases" with no defined dates or deliverables.
Who on the internal team will be able to operate this once you're no longer involved?
This is the question that makes consulting firms most uncomfortable when their business model depends on client dependency. A good answer names specific roles, describes what capabilities that team will have at the end of the project, and explains how knowledge transfer works.
If the answer is "that's what our support is for," you have a problem. Support is not autonomy.
How is project success measured, and who measures it?
Ask for success indicators to be written into the contract — not as aspirational objectives, but as verifiable metrics: process hours eliminated, error rate before and after, volume of queries resolved without human intervention.
A consulting firm that won't commit to concrete metrics generally has good reasons for avoiding them.
What happens if the agent fails in production?
AI systems fail. This is not a remote possibility — it is a statistical certainty. What separates a serious implementation from an amateur one is the response protocol.
Ask: Is there active monitoring? Who receives the alert? How quickly is it resolved? What manual process covers the gap in the meantime?
If there are no clear answers to these questions, you absorb the operational risk.
How much will it cost to keep this running each month?
AI infrastructure costs are not trivial. Language models, APIs, storage, orchestration — everything carries a recurring cost. Request an estimate of monthly operating costs before you sign, not after the system is in production.
A reasonable range for a mid-size company running two or three agents in production can fall between 800 and 3,000 euros per month in infrastructure, depending on volume. If nobody has mentioned this, treat it as a warning sign.
Do you have experience with companies our size?
Implementing AI at a 500-person company with a dedicated IT team is not the same as doing so at an 80-person company where the person responsible for technology also manages other areas. The methods, timelines, and risks are different.
Ask for concrete case studies — not generic testimonials: cases with context, the problem, the solution, and a measurable result.
What tools and platforms will you use, and why?
A consulting firm that always works with the same tools regardless of the problem isn't designing a solution for you — it's fitting your problem to what they already know how to do.
Ask why they chose those tools for your specific situation. The answer should include a comparison with alternatives and a justification grounded in your operational context.
How is access to our company's data managed?
Any AI agent operating on internal data — invoices, contracts, financial reports, customer data — requires a clear policy on data access, retention, and deletion.
Ask: Does the data pass through third-party servers? Which ones? Under what conditions? Is there a signed data processing agreement?
In Spain, this carries direct implications under the GDPR. This is not a technical detail — it is a legal requirement.
What does the price include, and what does it not include?
AI proposals typically carry a project price and a list of exclusions that surface later. Ask the contract to specify what is within scope and what generates additional costs.
Frequent areas of ambiguity: integrations with existing systems, scope changes during the project, team training, post-delivery support.
What is the plan if we decide not to continue with you after the project?
A consulting firm that builds structural dependency has no incentive to answer this question well. One that works with an enablement mindset does.
What your team should be able to do at the end of the project: access all the code, understand the architecture, modify the agents without external assistance, and scale the system independently.
If the answer includes phrases like "that would require a transition managed by us," you have the answer you needed.
An example to put the risk in perspective
A distribution company with 120 employees engaged a consulting firm to automate invoice reconciliation with suppliers. The project ran four months and cost 40,000 euros. When it concluded, the system worked — but only the external consultant knew how to modify it. When a key supplier changed their invoice format, the process broke. Reactivating it took three weeks of work and an additional 8,000 euros.
The problem wasn't the technology. It was that nobody asked the right questions before signing.
With a delivery model oriented toward autonomy, that same project could have left the internal team equipped to resolve that kind of incident in hours, without depending on anyone external.
Conclusion
These questions are not designed to distrust consulting firms. They are designed to separate those that work with rigor from those that work with well-produced presentations.
A consulting firm that answers all of these questions clearly — and is willing to commit those answers in writing — is a consulting firm worth working with.
If you'd like to review your specific situation before making a decision, you can complete the free diagnostic form. No immediate call, no commitment.
