We sell AI consulting, so treat this the way you'd treat a roofer's guide to choosing a roofer: useful, but check the incentives. Ours are simple — a shop that picks the wrong consultant, even a competitor, usually swears off the whole category, and that's bad for everyone who does this work honestly.
The five questions
1. "Will you come to our shop?"
You can't diagnose a shop over Zoom. Quoting problems live in filing cabinets and tribal knowledge; routing problems live on the floor. A consultant who won't visit is planning to sell you something generic. Anyone serving Northeast Ohio manufacturers should be physically in Northeast Ohio regularly — it's a fair filter.
2. "What's the smallest first project you'll sell us?"
The answer tells you their business model. "A discovery phase, then a roadmap, then implementation" means months of fees before anything works. The right first project is small, bounded, and testable — one workflow, a fixed price, a hard end date, and a decision at the end: scale it or stop. (Ours is 30 days and $4,500; anyone structurally similar passes this question, not just us.)
3. "Who does the work after you leave?"
If the answer isn't "someone on your team, whom we train," you're buying a dependency, not a capability. The test of a good pilot is that your people ran the tool during it — not watched a demo of it.
4. "What happens to our data?"
Your quotes, customer drawings, and pricing are the crown jewels. A competent consultant can say precisely which tools will see them, on what account plan, with model training turned off, and what leaves when they leave. Vague reassurance here is disqualifying — and if the consultant can't answer it for their own tools, they can't govern yours. (Our free AI Exposure Check shows you the questions worth asking internally, too.)
5. "What would make you tell us to stop?"
The most revealing one. AI genuinely doesn't fit every workflow — messy inputs, tiny volumes, or a process that's really a people problem. A consultant with no stop conditions has no honest ones. You want to hear something like: "If the pilot can't beat your current process by week three, we say so and you keep the write-up."
What it should cost
For a 10–200 person shop: a scoped pilot in the low thousands, ongoing help at a flat monthly rate you can cancel, and no platform purchases until something has already worked on your data. Six-figure proposals are for companies with IT departments. "Free" assessments that funnel into commission-based software sales cost the most of all — you just pay through the software bill.
Red flags, quickly
- A tool demo before they've asked a single question about your workflows.
- Reseller commissions on the software they recommend.
- ROI percentages promised before they've seen your data.
- No defined end point, success measure, or off-ramp.
- Case studies that can't be verified with a phone call. (We publish none yet for exactly this reason — we won't fake them, and we say so out loud.)
Put us through the five questions
First conversation is free, on the phone or at your shop — and question five gets answered in writing.
Contact UsOr the fastest way: call or text (216) 999-5765.