Before accepting OEM AI tools, a dealership should ask seven questions: what problem it solves, what data it reads or writes, where that data goes, who owns it, who approves it, what it costs later, and how to exit.
Last updated: November 3, 2026
It tends to start on a regional call. Somebody from the manufacturer’s side mentions an AI program they’re rolling out to dealers, the follow-up email names the vendor the brand certified and how much of the bill co-op will pick up if you enroll before the end of the quarter, and by Thursday your GM has heard from a friend two stores over that they already signed. Now saying no feels like the risky choice, which is a strange place to end up for a tool nobody at the store has actually looked at yet.
So the store signs, and I don’t blame anybody for that.
You’ll be selling that brand for a long time, part of the cost is covered, and the rep on the phone is someone you’d rather keep on your side. I’ve been with Helion for more than 20 years, though, and somewhere along the way I quit making exceptions based on who suggested a system. Anything that touches dealership data gets the same look from us, because a tool that shows up without a plan, and without anyone asking how the store really runs on a normal Tuesday, usually ends up making work for the very people it was supposed to help. Somebody has to own it. Somebody has to support it, and somebody has to know what it can see.
For the record, I’m not against manufacturer programs, and I’d expect a few of them to pay off nicely. What I’m describing is just the diligence you’d owe any vendor, the manufacturer’s pick included. Our managed AI services for dealerships start from governance before adoption, if you want the wider view of how we work. This post is narrower: the list I’d want answered before anybody signs.
Skeptical isn’t the same as hostile.
What Counts as a Manufacturer-Pushed AI Tool
A manufacturer-pushed AI tool is AI software that an automaker recommends, certifies, or helps pay for and offers to its franchised dealers through a program, usually built by an outside vendor the manufacturer approved and often tied to co-op funds, incentives, or an enrollment deadline.
What the tool does can be almost anything. I’ve seen chat for the website, lead follow-up that texts a customer back, service reminders, listing descriptions, and now and then something that reads repair orders. I won’t try to grade whether any of them sells more cars or books more appointments, partly because most customer-facing AI sits with the CRM and software vendors rather than an IT company like ours, and partly because it doesn’t change my part of the job. My part is the same whatever the tool does, and it comes down to three things, which are governance, data access, and ownership.
It’s worth separating this from the security standards automakers already set for their dealers, which our guide to OEM cybersecurity requirements for dealers walks through. Those standards are about how you protect your network. An AI program is something else, really, a new vendor knocking on that network and asking to be let in.
The logo on the program doesn’t change the review.
Why Saying Yes Feels Like the Safe Call
Part of it is that the manufacturer picked the vendor, so it feels like somebody already did the checking. Part of it is money, since an incentive makes the first year look close to free. And then there’s the deadline, which is usually what kills the review, because nobody wants to be the person who held up enrollment to ask about data retention.
I don’t think any one of those is unreasonable. Stack them up, though, and you get the pattern I run into most with AI right now: a store adopts new technology quickly, without proper planning and thought, and months later the cost turns up as a tool nobody opens, a login nobody remembers approving, or a monthly charge that lands on the CFO’s desk with no owner’s name next to it.
Industry research points the same direction. In Cox Automotive’s AI in Auto Retail Tracker, released in August 2026, 82% of dealers said they use AI and 69% expect it to drive sales and revenue growth, yet only 22% of AI users report seeing that growth so far, and about 1 in 3 either aren’t measuring AI’s impact or have no clarity on how they’re measuring it. Buying in is easy. Getting results takes a plan.
Then there’s the compliance angle, which people tend to forget when the offer comes from the manufacturer. A dealer that arranges financing falls under the FTC Safeguards Rule, and the rule asks it to take reasonable steps to select service providers that can protect customer information, to require those safeguards by contract, and to periodically assess each provider based on the risk it presents (16 CFR 314.4(f)). Nothing in there cares who recommended the vendor. If a tool handles your customers’ information for you, I’d treat it as your service provider and review it the way you would any other.
An endorsement is a reason to look closely, not a substitute for looking.
Seven Questions to Ask Before You Say Yes
I’d put these to the program contact and the vendor, in writing where you can, and keep every answer in one file. Some you’ll settle on a quick call. A couple will mean sitting down with the contract, which, in my experience, is where the real answers usually turn out to be.
1. What problem does it solve at this store, and how will we know it worked?
Forget the program brochure for a minute and start with your own store. Maybe your web leads wait until morning for a reply, or your service lane can’t keep up with appointment calls on Mondays. Whatever it is, name it in your own words, pick the one number that would tell you the tool fixed it, and write down where that number sits today, before anything gets switched on.
If nobody at the store can say what the tool is for, it isn’t ready, however good the incentive looks. Our post on why dealership AI projects stall goes further into writing that plan.
2. Which of our systems does it read or write, and through what access?
This is the one I care about most. As a CISSP, access is the first thing I check on any new tool, and AI raises the stakes, because these tools are only useful once they’re plugged into something real, whether that’s the DMS, the CRM, the phone system, or the customer records behind your deals. So ask exactly which systems it touches, whether it only reads from them or can also write back, and how the connection gets made. A named integration running under its own limited account is fine. If the vendor asks for an administrator login so they can “get everything set up,” I’d stop right there until somebody explains, in writing, why they need it.
Here’s the part people underestimate. An AI tool will show people whatever it’s been given access to, so if you hand it everything, it can surface everything to anyone who happens to ask the right question. The safer setup follows least privilege, meaning the tool gets the minimum access its job needs and not a bit more. We wrote about least privilege for dealership AI on its own, since it applies to every AI tool in the building and not only this one.
3. Where does our data go, and does it train anyone’s model?
Get it in writing. Ask where your data is stored, whether it passes to other companies the vendor uses, how long it’s kept, and whether your customers’ information or your staff’s conversations can be used to train or improve a model, the vendor’s or anyone else’s. Ask who owns what the tool produces from your data, and whether you can take it with you.
Then ask the follow-up. What happens if those terms change? In February 2024, the FTC warned AI companies that quietly changing terms of service or privacy policies to allow AI training on data already collected could be unfair or deceptive. That warning was aimed at the companies. For a dealer, the lesson is simpler: pin down the terms you’re agreeing to today, and make sure the contract says you’ll be told before they change. If the training question is new to you, our explainer on whether AI tools train on your data walks through how the settings work.
4. Who at the store owns it, and who supports it?
A person. By name.
Not “the manufacturer,” and not “the vendor.” Ownership means someone at the store decides who gets access, checks whether the tool is doing what it was bought to do, and makes the call to renew or cancel. Support is a separate question, and it’s the one that gets missed. When the tool stops working on a busy Saturday, does the call go to the vendor, to the program’s help line, or to your IT team? Does anyone at the store know which?
Supportability is part of the price. Always. A tool nobody can support costs more than its invoice says.
5. Does it fit the AI platforms we’ve already approved, and who signs off?
If your store keeps an approved list of AI tools, this one goes through it like anything else. No list yet? Start one. This tool is a good excuse. At a dealer group that usually means an executive committee that approves every AI tool before it rolls out; at a smaller store, the dealer principal or the CFO is the approver. We laid out how that works in our guide to setting up a dealership AI governance committee, and if you’re earlier than that, our overview of AI governance for dealerships is the place to begin.
Think about standardization while you’re at it. Every extra AI tool brings its own login, its own copy of customer data, and its own support path. For a multi-rooftop group, decide whether the tool runs at every store or none of them, because a different setup at each rooftop is how a simple environment stops being simple.
6. What does it cost once the incentive period ends?
Co-op funds and launch incentives cover a window. Ask what the store pays after it closes, per rooftop and per user, what renews on its own, and how much notice you get before the price changes. Then add what isn’t on the vendor’s invoice: someone’s time to manage the tool, the integration work, the training, and the review at renewal. Multiply by the number of rooftops if you’re a group. It adds up fast.
Partly paid for is still a price.
7. How do we get out?
Ask this one early. Every contract should answer it before anyone signs. How much notice ends it? Does leaving early cost you any of the incentive? When it ends, will the vendor return your data, delete it, and confirm the deletion in writing? And on the last day, who at the store shuts off the tool’s connections to the DMS, the CRM, and everything else it touched? Our AI vendor risk assessment for dealerships covers offboarding in more depth.
A tool you can’t leave cleanly was never really a trial.
The Seven Questions on One Page
Print this, fill in the last column, and keep it with the contract.
| Question | Who should answer it | What to get in writing |
|---|---|---|
| What problem does it solve, and how will we measure it? | The manager whose department will use it | The problem, the measure and today’s number |
| Which systems does it read or write, and through what access? | The vendor, checked by your IT team | Each system, read or write, and the account it uses |
| Where does our data go, and does it train a model? | The vendor | Storage, retention, training use, and notice of changes |
| Who owns it, and who supports it? | Your leadership team | One owner by name and the support path |
| Does it fit our approved AI platforms, and who signs off? | The AI committee, or the dealer principal or CFO | The approval, dated |
| What does it cost after the incentive period? | The program contact and the vendor | Price per rooftop and user, renewal terms, notice |
| How do we get out? | The vendor’s contract | Notice period, data return and deletion, access shut-off |
What We Hear When Stores Ask These Questions
Most of the pushback is the same four lines.
“The manufacturer already vetted it.” They vetted it for the program’s goals. Your data, your access controls, and the way your store runs are still yours to check, and the Safeguards Rule puts vendor oversight on the dealer.
“If we miss the deadline, we lose the money.” Maybe. Ask whether enrolling and connecting your systems have to happen on the same day. If they don’t, you can hold your place in the program and still do the review first.
“It’s free.” For now. See question 6.
And the one I hear most from groups: “Our IT person already looked at it.” Good. Ask what they looked at. Checking that the tool installs and runs is a different job from reviewing its access, its data terms, and its exit, and a busy IT manager handed a deadline will usually do the first one and assume someone else did the rest.
Ask what was checked, not whether.
Where Helion Fits
We don’t sell manufacturer AI programs, and we don’t judge whether a customer-facing tool sells more cars. What we do is the review around it: which systems the tool reaches and through what access, what the contract says about your data, who owns it at the store, and whether it fits the AI platforms and governance you already have. That part is our job. With 2,000-plus dealerships under management, we look at a lot of vendor connections, and the questions above are the ones that keep coming back. For groups that want it handled as an ongoing program, that review is part of our managed AI work.
Governance first. Then the tool.
What Dealers Ask About Manufacturer AI Programs
Can a store turn down a manufacturer AI program without hurting the relationship?
The manufacturer chose the vendor. Doesn’t that make it their responsibility?
We signed up months ago. Too late?
Who should be in the room for this review?
Say Yes on Paper, Not Under Pressure
The manufacturer wants the program to work, and so do you. These seven questions are how you make sure it works at your store, with your data, run by someone who owns it, and how you avoid finding out a year from now that nobody can say what the tool can see. Keep the answers with your vendor files and ask them again at every renewal. If you’d like a second set of eyes on where your environment stands before the next program arrives, our complimentary IT and cybersecurity assessment looks at how your systems are set up and who can reach what, which is the ground any new AI tool will stand on.
Simpler, more secure, and easier to support is still the test, whoever is offering the tool.