12 questions to ask before hiring an AI development company

MIT found 95% of enterprise generative-AI pilots deliver no measurable P&L impact. The gap is rarely the model — it's integration, ownership and measurement. The right questions up front tell you which side of that 95% a partner will land you on.

On capability and fit

1.What have you shipped to production — not demoed? Ask for live AI systems with real users. (Here's ours.)
2.Is this built on your own architecture, or a thin wrapper? Understand what's "under the hood" — proprietary, open-source, or a reseller of someone's API.
3.How will you measure success in business terms? If the answer is "accuracy" rather than a P&L line, push harder. (Why ROI is the real skill.)

1.Who owns the model, the prompts and the outputs? It should be you, in writing.
2.What happens to our data — during and after? Contracts should spell out exactly what the vendor can and can't do with your data.
3.How do you handle security and compliance? Encryption in transit and at rest, access controls, and your industry's regime (HIPAA, SOC 2, GDPR).
4.How do you prevent hallucinations and handle errors in production? Evaluation, guardrails, human-in-the-loop where it matters.

1.What's the full running cost — including the maintenance tail? Inference, monitoring and prompt upkeep are ongoing. (The AI cost curve.)
2.Can we start with a paid pilot? A small proof-of-concept is the best "try before you buy at scale" there is.
3.Who specifically works on our account, and how senior are they?

1.What happens if we want to bring this in-house later? A confident partner makes that easy.
2.Show me a project that went wrong — what did you do? How a team handles failure tells you more than any case study.

A partner who answers all twelve crisply is rare — and worth far more than the one with the slickest demo.

If you're scoping an AI build, bring us these questions — we'd rather you ask them of everyone you're considering, including us.

Written by ivector