This website uses cookies

Read our Privacy policy and Terms of use for more information.

Key Takeaways:

  • Every AI tool in your dealership is only as good as the data sitting underneath it, and most dealership data starts out messier than leadership realizes.

  • The information that actually makes AI useful — your SOPs, your training manuals, your service department know-how — usually lives outside any system an AI can currently reach.

  • The dealerships winning with AI a year from now will be the ones using it to make their existing team more effective, not the ones chasing replacement.

At the AutoIndustry.ai Summit 2026, a panel moderated by Michael Cirillo pulled three operators away from the demo floor to talk about the layer underneath every AI pitch: the data foundation. Gary Ladue of Modern Automotive, Daniel Kim of Barnes Crossing Auto Group, and Kyle Mountsier of More Than Cars and Auto Genius walked through it as a climb, not a single decision — five rungs, each one unlocking the next.

Rung 1: Your DMS is Dirtier Than You Think

Kim's first move at Barnes Crossing was simple: try to access the dealership's own data. Years of history sat there, buried under customers who'd bought elsewhere years earlier and vehicles still attached to owners who'd long since moved on.

Kim put the moment plainly:

"Once you apply hygiene to the data, you can see how much gets corrected — that's pretty eye opening."

Rung 2: The Good Stuff Still Sits Outside Any System

Mountsier's sharpest point of the session: most AI vendors have already mapped a dealership's structured data. The valuable layer sits somewhere else entirely.

"The problem is the non-structured data that actually holds the operational context that's necessary to build on top of this."

That means SOPs, training documentation, and the specific way a service advisor answers a tough question — the stuff that lives in people's heads, not in a database.

Rung 3: Get Your Vendors Talking

Every AI vendor collects gold on your behalf. Most of it never comes home.

Kim named the exact failure point:

"If it's living outside of your data source, like a CDP, then it's not as effective."

And the reason so much slips through: 80–90% of leads still arrive as bare ADF XML, stripped of the actual conversation that produced them.

Rung 4: Turn a Clean List Into Real Reach

Ladue's team runs Liminal across 17 rooftops, and once the org understood what the platform could do, the requests started writing themselves — audience after audience, one idea leading to the next.

A 50-car store and a 400-car store fill those audiences at completely different speeds, and Ladue's team is still learning the pace store by store.

Rung 5: Build What You Can Defend

AI coding tools have made building genuinely fast. Kim built a working trade-in widget in 45 minutes during the event itself. The caution that comes with that speed:

"That's a significant barrier for me, and that's security risk."

His estimate: about 20% of dealerships today have the technical maturity to build safely in production. Everything customer-facing still needs a real answer on encryption, compliance, and where the data lives.

The Top of the Ladder

Every panelist landed on the same place. Ladue's version of it closed the session:

"We're not looking to replace people. We're looking to enhance the output of what we already have."

What This Means for Any Dealership Right Now

  1. Run a real hygiene audit on your DMS data. Expect outdated ownership records and stale entries in numbers that surprise you.

  2. Document your operational knowledge, not just your inventory. Your SOPs and training manuals are what make AI genuinely useful inside your store.

  3. Push every AI vendor for a feedback loop. Insist any AI tool routes what it collects directly back into your central CDP.

  4. Weigh build versus buy honestly. Buy core systems like CRM and DMS. Build custom tools around specific workflows once your team understands the security requirements involved.

  5. Track training speed as a signal. A dealership with well-structured operational data trains a new employee faster, and that same structure is exactly what helps AI succeed too.

Reply

Avatar

or to participate

Keep Reading