What to Know Before Hiring an AI Vendor

The Market is Noisy

Every consulting firm now has an “AI practice.” Every dev shop has a “machine learning team.” The barrier to claiming AI expertise is a website and a pitch deck. The barrier to delivering production AI systems is considerably higher.

Five Questions That Matter

1. Will I own the code?

If the answer involves licensing fees, proprietary platforms, or ongoing dependencies, walk away. You should own every line of code, every model, every piece of documentation from day one.

2. Where does the work happen?

If the work happens in the vendor's environment, on their infrastructure, in their repo — you are building a dependency, not a capability.

3. How fast will I see production results?

“Discovery phase” is often a euphemism for “we don't know what to build yet.” A vendor with domain expertise should scope a production-ready workflow in the first week.

4. What happens when the engagement ends?

Can your team maintain and extend the system independently? Is there documentation? Is there knowledge transfer? Or are you calling the vendor every time something breaks?

5. Do they know my industry?

AI engineers who have never seen a rent roll, a claim form, or a lease agreement will spend months learning your domain before delivering anything useful. Domain expertise is the difference between a 30-day deployment and a 6-month exercise.

Red Flags

• 200-page strategy documents. If the deliverable is a PDF, not a production system, you are paying for paper.
• Hourly billing. Misaligned incentives. You want speed; they want hours.
• No case studies with measurable outcomes. Testimonials are not the same as metrics.
• “We can do anything.” The best vendors have a clear focus. Generalists spread thin; specialists go deep.

What Good Looks Like

A vendor who scopes tightly, ships weekly, embeds in your team, and gives you ownership of everything they build. Predictable pricing. Measurable results. Month-to-month commitment because the work speaks for itself.