← All postsJuly 29, 2026

Questions to Ask Before Hiring Any AI Vendor

The Federal Trade Commission keeps catching AI vendors selling something other than what they claimed. In May 2026, it required Cox Media Group and two smaller marketing firms to pay nearly $1 million combined, after finding their "Active Listening" service — pitched as AI that detected relevant conversations picked up by smart devices to target local ads — didn't listen to anything at all. It was reselling data-broker email lists at a markup. Separately, in March 2026, the FTC banned a company called Air AI from ever marketing a business opportunity again after finding it misled entrepreneurs with inflated earnings and growth claims; the settlement's headline $18 million judgment was largely suspended, with the operators actually required to pay $50,000, based on their inability to pay the rest. Legal analysts tracking the pattern count this as roughly the FTC's thirteenth "AI-washing" case since 2024 — a product wrapping a basic capability, or none at all, in "proprietary AI" language and selling it at a markup.

You don't need to become a regulator to avoid becoming case number fourteen. A few plain questions, asked before signing anything, do most of the work.

What to ask

  • What specifically does the product do, in one sentence, without the word "AI" in it? If they can't answer that, there's usually nothing there — or nothing worth the price.
  • What happens if I stop paying? Do you keep your data and configuration, or does it all live inside their platform permanently?
  • Can I see it work on my own data, before I commit? A live demo on a real example beats a slide deck every time.
  • Who's responsible when it's wrong? AI tools make mistakes — ask what the fallback is, and who owns the outcome when it fails.

Red flags

  • Guaranteed results — revenue, leads, time saved — stated as fact rather than a range.
  • Pressure to sign before you've had a chance to test anything.
  • Vague answers about what's actually running underneath the marketing language.

None of this means AI tools are bad — plenty are genuinely useful. It means the burden of proof belongs on the vendor, not on you.

Sources: