In just a few years, the corporate AI market has filled up with providers of very uneven quality. Some build tools that genuinely work; others mainly sell a promise. Here are the questions that help you tell the two apart before you sign.
"Do you use what you sell yourself?"
This is the most revealing question. A provider who runs their own tools daily, facing their own constraints, understands the subject differently from one who only sells what others have built for them.
"What happens if the project concludes that nothing should be built?"
A genuinely independent audit can conclude that no tool is needed. If a provider cannot accept that outcome, it's a sign their audit is just a disguised sales step, not an honest assessment.
"Who takes care of the tool after delivery?"
A tool delivered without a maintenance plan degrades over time. A reliable provider explains clearly what happens after delivery: who fixes bugs, who adjusts the tool as the business evolves, and at what price.
"How do you handle GDPR and AI Act compliance?"
A provider who cannot answer this question precisely exposes the business to a risk it will discover later, at the worst possible moment. The answer must be concrete, not a vague line about "compliance by default."
"Does the price include everything, or should we expect hidden costs?"
Every stage (audit, build, maintenance) should have a defined price before work begins. A provider who cannot give a precise range, even a broad one, has probably not done this type of project enough times to estimate it properly.
Delivers recommendations, not tools: that is the limit of a consulting firm. Billed by time spent rather than by results: that is the limit of a generalist services company.
