7 Red Flags When Hiring an AI Automation Vendor

The seven flags

1. Unstatable pricing. If a vendor cannot tell you what one workflow costs to build and to run, they are pricing by what you will tolerate. 2. No written guarantee. Diagnostic work should carry a money-back condition in writing; delivery work should carry acceptance criteria. 3. Demo-ware. A demo on the vendor’s sample data proves nothing — insist on a scoped test against your actual tools before a large commitment.

4. Vendor-hosted lock-in. If the workflows run on infrastructure you cannot access, the price of leaving is rebuilding everything. Prefer vendors who build inside your accounts so you own the result. 5. No named humans. If you cannot name the person who will build your system, you are buying a queue, not a service. 6. Boil-the-ocean scope. Serious vendors sequence: highest-value workflow first, prove it, then expand. Anyone proposing a 12-workflow transformation on day one is selling a retainer, not results. 7. Unfalsifiable case studies. Real proof survives follow-up questions; composites and illustrations are legitimate only when clearly labeled as such.

What good looks like instead

A trustworthy engagement starts small and specific: a fixed-price assessment with a written guarantee, a per-workflow build price, operations pricing you can cancel, systems built inside accounts you own, and a named builder. Ask every vendor the same question — "what happens if this doesn’t work?" — and hire the one with a concrete answer.