Every vendor demo looks impressive. The screen is clean, the numbers appear instantly, and the salesperson is fluent in AI. None of that tells you whether the tool will help your dealership or quietly cost you money for two years.

This is a buyer's checklist. It is built around one principle: a good AI tool can show its work. If it can't, you are not buying intelligence - you are buying confidence you can't verify.

Use the questions below in your next vendor conversation. The way a vendor reacts to them tells you almost as much as the answers.

The core test: can it show its work?

Before any feature list, ask one thing. When this tool gives me an output, can I see why?

A valuation, a price suggestion, a lead score - each is a decision dressed as a number. If the tool can point to the evidence behind it, you can check it, defend it and learn from it. If it can't, you are being asked to trust a stranger's model with your margin.

Everything that follows is a way of pressure-testing that one question.

1. Does it show its sources?

Ask the vendor to produce a single output - a valuation is ideal - and then ask: what is this based on?

A credible tool answers concretely. For a valuation, that means the comparable vehicles used, how recent they are, and the main factors driving the figure. For a lead score, the signals that moved it.

Green flag: Sources are visible by default, not hidden behind a support ticket.

Red flag: "Our proprietary algorithm." That phrase is doing the work of an answer without being one. Proprietary is fine. Unexplained is not.

Why it matters at the desk: when a customer challenges a trade-in figure, "the system says so" loses the deal. "Here are six similar cars that sold in the last three weeks" wins it.

2. Does it expose confidence?

Every estimate is uncertain. The honest tools tell you how uncertain.

Ask: does this give me a range and a sense of how sure it is, or just one number?

A common, liquid car with lots of recent data should price tightly. A rare specification with thin data should come back with a wide range - and the tool should say so. A system that returns a single confident number for every car, no matter how unusual, is not more accurate. It is hiding its own uncertainty from you.

Green flag: Confidence intervals or ranges, with narrower bands where the data is strong and wider where it's thin.

Red flag: One hard number, always, with no sense of how much to trust it.

This is the difference between a tool that helps you decide and one that just decides for you and hopes you don't notice when it's wrong.

3. Does it log outcomes?

This is the question most buyers never ask, and it is the one that separates serious tools from demos.

Ask: does the system record what it predicted, then compare it against what actually happened?

An AI valuation tool should log the value it estimated and later check it against the price the car actually achieved. A days-to-sell model should track its prediction against reality. This is outcome logging, and it does two things. It lets you measure whether the tool is actually accurate on your stock, not on a vendor's slide. And it lets the tool improve over time, because it can learn from its own misses.

Green flag: The vendor can show you accuracy tracked over time, on real outcomes, and ideally broken down by segment.

Red flag: No accuracy tracking at all. If a vendor cannot tell you how often their tool is right, assume they don't know - and that they have no mechanism to get better.

A tool that doesn't measure itself cannot improve. You would be buying today's accuracy forever, with no compounding.

4. Who owns the data?

Your dealership generates valuable data: every appraisal, every sale, every outcome. When you use an AI tool, that data flows in. The question is what happens to it, and whether you can ever get it back.

Ask: if I leave in two years, do I keep my data, and in what format?

Green flag: Your data stays yours, exportable in open, standard formats. You can walk away with your history intact.

Red flag: Your data becomes effectively the vendor's, locked in a format you cannot use elsewhere. That is not a partnership. That is a hostage situation with a monthly invoice.

Data ownership is the quiet decision that determines how much leverage you have at every renewal. Decide it on day one, not at the exit.

5. Is it auditable?

When something goes wrong - a wildly off valuation, a strange price suggestion - can you find out why?

Ask: if this tool makes a bad call, can I trace what happened?

Auditability means you can reconstruct a decision after the fact: what data went in, what the tool concluded, and why. Without it, every error is a mystery, and you cannot tell a one-off glitch from a systematic problem that is bleeding margin across your whole lot.

Green flag: A clear trail from input to output that you can inspect.

Red flag: Errors that vanish, with no way to understand or reproduce them.

6. Does it run alongside what you have?

Few dealers can or should rip out their core systems to adopt one AI tool. The lower-risk path is augmentation: the tool runs alongside your existing setup, adds value, and earns expansion.

Ask: can I trial this next to my current process without a forklift migration?

Green flag: It integrates with what you run and proves itself before you commit deeply.

Red flag: All-or-nothing replacement before you have seen a single real result.

Augment-first is not just gentler. It is how you keep the trial honest - you can compare the tool's output against your existing baseline, in the open.

7. Where does your data live, and under whose rules?

For European dealers, data sovereignty is not a checkbox. It is regulation and it is trust. Customer and vehicle data carries obligations under EU law, and the EU Data Act is reshaping who can access and move data generated by connected products.

Ask: where is my data stored, under which jurisdiction, and does this respect EU data rules?

Green flag: Clear answers on data residency and EU compliance, framed as a feature rather than a reluctant disclosure.

Red flag: Vagueness, or storage and terms that leave you exposed.

The red-flag summary

Walk away, or push hard, when you see:

  • Unexplained outputs. "Trust the algorithm" with no sources.
  • No confidence signal. One hard number for every car, never a range.
  • No accuracy tracking. The vendor cannot tell you how often it's right.
  • Data lock-in. You can't leave with your own history in a usable format.
  • No audit trail. Errors that can't be traced.
  • Forced replacement. Big commitment before any proof on your real data.

Any single red flag is a conversation. Several together is a decision.

The standard to demand

The market has been burned before by black-box pricing tools that gave confident numbers and no way to check them. The lesson is not "avoid AI." It is "demand AI that can show its work."

The standard to hold every vendor to is simple: transparent, outcome-logged, auditable, and built on data you own. A tool that meets that bar gets better the longer you use it and keeps you in control. A tool that doesn't is a black box you are paying to trust.

When you can see the sources, see the confidence, and see the track record, adoption stops being a leap of faith. It becomes a measured decision - which is exactly what buying anything for your dealership should be.

VehIQ is built to pass this checklist by design: shown sources, exposed confidence intervals, outcome-logged AI, data you own in open formats, and an augment-first model that runs alongside your existing DMS. EU data sovereignty by default. The trust layer for European automotive.