The One Question to Ask Every AI Vendor

Before you sign any AI contract, ask the vendor one question: "What data are you running on, and do I own it?" Write the answer down. Vendors who can't answer clearly are renting you their intelligence, and renting stops the day you stop paying.

Feature lists, demos, and ROI decks are all designed environments. This question is not, and the unrehearsed thirty seconds after you ask it will tell you more than the rest of the meeting combined.

Why this question and not fifty others

Every dealership AI evaluation drowns in the wrong questions. Which models do you use? How accurate is it? What integrations do you have? Those answers are rehearsed, and mostly interchangeable across vendors anyway.

The data question is different because it locates the asset. AI capability is becoming a commodity; every vendor rents roughly the same frontier models. The durable value in any AI deployment is the data it runs on and who holds it when the relationship ends. Asking about data ownership is really asking: at the end of this contract, do I own an asset, or do I own a receipt?

It also splits the market cleanly. Half the AI being sold to dealers is outward-facing tooling running on generic models, fine as a utility, worthless as a moat. The other half claims to be intelligence. This question is how you find out which one is across the table.

Reading the answers

"We run on your data, in a layer you own, and everything we enrich stays yours." The right answer. Verify it in the contract: full export rights including enriched fields, no fees, defined turnaround. If the paper matches the pitch, this vendor is building your asset.

"We run on our proprietary models trained across thousands of dealerships." Honest, and clarifying. You're buying aggregate intelligence, the same product your competitor can buy Tuesday. Sometimes worth having. Never confuse it with advantage, and never pay moat prices for utility products.

"Your data stays in your instance, but the enrichments are proprietary." The half-answer to catch. The scores, predictions, and segments built from your customers stay with the vendor, which means the longer you use the product, the more of your intelligence lives in their building. Switching cost is the product.

"That's a great question, let me get back to you." They will return with marketing language. The fact that the answer required engineering and legal tells you the ownership story was never designed for your benefit.

A confident answer that won't go in the contract. The most common outcome. Whatever was said in the room, the agreement says otherwise or says nothing. The contract is the answer. Everything else was the demo.

The two follow-ups

Once the first answer lands, two more questions finish the map.

"If we cancel, what do we walk away with, in what format, and how fast?" Renting intelligence stops when you stop paying. This question prices the stop. Flat files in 60 days for a fee is a very different exit than full structured export on demand.

"Does our data improve products you sell to other dealers?" This surfaces the extraction model: your transactions training the benchmarks and algorithms sold across the industry, including to the store across town. If the answer is yes, that's not automatically a dealbreaker, but it is leverage. You are a supplier as well as a customer, and suppliers negotiate.

Make it standard procedure

Run the question retroactively, this week, on every AI tool you already pay for. Interview every AI vendor on the list, write down the answers, and build the grid: tool, data source, ownership, exit terms. Most stores discover they are renting their entire intelligence layer, paying $150K to $400K a year in vendor fees for the privilege.

Then make the question the gate for everything new. Owned intelligence compounds every day you operate. Rented intelligence is a subscription with a memory you'll never get back. One question, asked early and verified in writing, is how you tell which one just walked into your conference room.

FAQ

What is the most important question to ask an AI vendor? "What data are you running on, and do I own it?" The answer reveals whether you're building an owned asset or renting intelligence.

What is the right answer from an AI vendor? That the tool runs on your data, in a layer you own, with all enrichments included in full, free, fast export rights, confirmed in the contract.

What are the warning signs in vendor answers? Proprietary enrichments built on your customers, aggregate-model-only products priced as moats, vague answers requiring follow-up, and confident claims the contract won't back.

What follow-up questions should I ask? "If we cancel, what do we walk away with, in what format, how fast?" and "Does our data improve products you sell to other dealers?"

Should I ask existing vendors this question? Yes. Audit every AI tool you currently pay for and build a grid of data sources, ownership, and exit terms. It becomes leverage at every renewal.

Todd Smith is the Founder and CEO of QoreAI and the author of The Intelligent Dealership: How AI and Data Transform Automotive Retail.