Resale Royalty: How to Spot Thin AI Products Wearing Enterprise Price Tags
There is a particular species of software company flourishing in the current market that deserves a name. Call it the wrapper-in-a-suit: a product that does little more than relay your prompts to someone else's foundation model, apply a light coat of branding, and invoice you as though proprietary intelligence were involved. The suit is excellent. The substance underneath is frequently a single API call.
To be clear, wrappers are not inherently illegitimate. A thoughtfully built interface on top of a general-purpose model can deliver genuine value — that is, after all, what most of the application layer looks like right now. The problem is pricing and posture. When a company charges enterprise rates for what is functionally a pass-through, while concealing the pass-through, procurement is being defrauded of leverage. Here are five signs you are looking at one.
One: The Accuracy Claim Has No Denominator
'Ninety-five percent accurate' is not a claim; it is a vibe. Accurate on what dataset, measured how, benchmarked against what baseline? Genuine AI vendors are proud of their evaluation methodology and will hand over the harness. Wrapper vendors rely on yours never existing.
Ask this in the demo: 'Which dataset produced that number, and can we run our own test set through your system before contracting?' Watch how quickly the room changes temperature. A vendor that resists an independent evaluation on your data is telling you something important.
Two: The Demo Avoids Your Data Like a Debt Collector
Every serious AI product demo should involve your documents, your queries, your edge cases. The wrapper-in-a-suit prefers canned examples, because canned examples were curated precisely where the underlying model performs well. Request a live session on a sample of your messiest real-world data. Thin products fail loudly in that setting; substantial ones fail interestingly, and explain why.
Three: Data Lineage Is a Shrug
Where does your data go when it enters the product? Which model processes it, in which region, with what retention? A vendor that cannot produce a clear data flow diagram — or that buries the answer in a forty-page DPA — is either reselling infrastructure it does not fully understand or hoping you will not read the appendix. Either way, your security review should treat the ambiguity as disqualifying. American buyers, particularly in regulated sectors like healthcare and financial services, have more leverage here than they realize. Use it.
Four: The Roadmap Is Someone Else's Roadmap
Ask what happens to the product when the underlying foundation model releases its next version. A company with genuine engineering depth will describe migration work, evaluation reruns, and prompt refactoring. A wrapper will describe an automatic upgrade — because there is nothing between the model and the customer to migrate. The absence of an integration layer to maintain is not efficiency. It is evidence.
Five: The Contract Is Confident About Vagueness
Red flags cluster in the paperwork. Accuracy or uptime commitments phrased as aspirations. No service-level definition for model quality, only availability. Broad indemnification language that evaporates on inspection. And the quiet one: pricing that scales with your usage of a resource the vendor obtains at wholesale rates, with no disclosure of the margin structure. You are entitled to know what you are paying a markup on.
The Bottom Line for Buyers
The wrapper-in-a-suit thrives on asymmetry — of information, of urgency, of FOMO. The countermeasure is procedural, not heroic. Demand evaluation on your data. Demand lineage documentation. Demand to understand the margin. Vendors with real substance will welcome the scrutiny, because scrutiny is precisely what they have built to withstand. The thin ones will tell you everything you need to know by how much they resist.
