How to Spot an AI Snake-Oil Pitch Before You Sign

There is more money being made right now selling AI than being made using it.
Some of that is honest work. A lot of it is a pitch dressed in the vocabulary of inevitability — a polished demo, a confident number, and a contract that needs signing before the quarter closes. For a small nonprofit or a family-owned business, the cost of not seeing the difference is a platform you'll shelve in six months and money that should have stayed in your program budget.
You don't need a technical background to protect yourself. Most snake-oil pitches share the same handful of tells, and once you can name them, they're hard to un-see.
Here is the checklist we run for clients — and, honestly, the one we run on our own industry.

Red Flag 1: ROI Numbers With No Source

"Organizations using our platform see a 40% increase in productivity."
Ask one question: whose organizations, measured how, over what period? A real number comes with a real study, a named comparison, and a method you can inspect. A snake-oil number comes from a slide deck and evaporates the moment you push on it.
If a vendor can't tell you where a statistic came from, treat it as marketing, not evidence. The burden is on them to source the claim, not on you to disprove it.

Red Flag 2: Manufactured Urgency

"Prices go up next month." "We only onboard a few clients per quarter." "Your competitors are already doing this."
Urgency is the oldest sales tool there is, and it exists to move you past the part of your brain that asks good questions. A tool that genuinely fits your organization will still fit it next month. The pressure to sign now is almost never about your calendar — it's about theirs.
The competitor line deserves special suspicion. "You'll fall behind" is unfalsifiable by design. Push it into something concrete — which competitor, doing what, with what result? — and it usually dissolves.

Red Flag 3: Demo-Only Proof

The demo is always flawless. It runs on the vendor's data, with the vendor's examples, in the vendor's hands.
The real question is what happens on your data, with your messy records, run by your staff on a Tuesday. Ask for a pilot on your own material before you commit to anything. A confident vendor will welcome it. A snake-oil pitch will explain why that's not possible yet — and that explanation is your answer.

Red Flag 4: No Straight Answer on Your Data

This is the one that matters most, and the one non-technical buyers most often skip.
Ask, plainly: Where does our data go? Who can see it? Is it used to train your model? Can we get it back and delete it?
A trustworthy vendor answers these in specifics without flinching. A snake-oil pitch gets vague, reassuring, and non-committal — "it's all secure," "don't worry about that." For a nonprofit holding donor records or a business holding customer information, a vendor who won't answer the data question clearly has already told you something important.

Red Flag 5: The Recommendation Never Changes

Watch what happens across several conversations. Does every problem you describe somehow require their product? Is the answer always "buy the thing," regardless of what you actually brought them?
An advisor whose recommendation is fixed before they understand your situation isn't diagnosing — they're selling. The tell isn't that they recommend their product. It's that they were always going to.
This is the conflict we built Breakwater to remove on purpose. We don't take commissions, resell software, or earn anything based on which tool you pick. That's not a virtue we're claiming — it's a structure that lets us say "wait" or "do nothing" when that's the honest answer, without it costing us anything. When someone's payday depends on your yes, their advice and their invoice point the same direction. Know which one you're getting.

Red Flag 6: No Answer for "What Happens When It's Wrong"

AI tools are wrong sometimes. That isn't disqualifying — pretending otherwise is.
Ask what happens when the system produces a bad output. Who checks it? How would you know? Can you audit why it did what it did? A serious vendor has thought about failure and has a human-review answer ready. A snake-oil pitch treats the question as rude, because "it just works" is the whole product.
Griffin's research is on exactly this — where intelligent systems fail and how to make them auditable rather than take them on faith. The organizations that get burned are rarely the ones that asked too many questions about failure.

The One Question That Cuts Through Most of It

If you remember nothing else, ask this:
"What would have to be true for you to tell me not to buy this?"
An honest advisor can answer immediately, because there are real conditions under which their tool is the wrong call and they know what they are. A snake-oil pitch cannot answer it at all — because in their model, there is no such condition. There never was.

Where We Come In

You shouldn't have to become an AI expert to avoid being sold one. That's the entire reason independent advice exists.
Our Industry Opportunity Scan does this work for you: independent research into what's actually shifting in your field, which tools are real, which are noise, and where — if anywhere — the opportunity is worth pursuing. No product to sell you at the end of it, because we don't have one. Just a clear read on what's worth your time and what's an expensive pitch.
Fifty years of watching technologies get sold as revolutions taught Barry the pattern. The names change; the tells don't.
If the answer is always "buy the thing," you already have your second opinion.
Learn more about the Industry Opportunity Scan or start a conversation.