A dealer told me last month he had “AI follow-up” running in the BDC. Then we pulled 20 lost leads. Seven had bad vehicle status. Four got texts on cars already sold. Two customers were marked duplicate because the spouse had bought a car three years ago. One lead had 14 automated touches and zero human notes. That store didn’t have an AI problem. It had a plumbing problem.
That’s the part getting skipped in a lot of vendor demos right now. AI sounds clean on a conference stage: smarter pricing, better customer engagement, automated service retention, faster trade acquisition, less admin work. I buy the general direction. I’ve seen stores use automation to pick up real time, especially around appointment confirmation, equity outreach, and service-lane acquisition. But I’ve also seen automation make a mediocre process louder.
CBT News recently framed it well: AI is only as strong as the infrastructure underneath it. I’d argue that is the most important dealer-tech sentence of the year, because most stores are still trying to bolt automation onto systems that were never cleaned up after the last three tool changes.
Your AI Does Not Know What Your Store “Meant”
Humans forgive messy dealership data all day long. A good used car manager knows the Tacoma showing “available” in the CRM is actually waiting on a bumper. A service advisor knows Mrs. Henson prefers text even though the profile says home phone. The BDC rep knows the “internet lead” is really a repeat customer who got annoyed when sales didn’t call back after her oil change.
AI does not know any of that unless the store’s systems say it clearly. It reads fields, timestamps, statuses, rules, templates, and behavior. If those are wrong, stale, duplicated, or half-used, the machine will act with confidence on bad information.
This is where operators need to get more honest. Most dealership data isn’t bad because people are lazy. It’s bad because the store grew in layers. A DMS migration here. A CRM change there. A new scheduling tool. A new chat provider. A new service retention platform. Three BDC managers with three different naming conventions. Before long, the store has 11 versions of “appointment set” and nobody wants to touch the rules because something might break.
The Four Layers That Decide Whether Automation Works
Here’s the framework I use when I’m looking at a store’s readiness for heavier automation. Not fancy. Just practical.
- Data hygiene: Are customer records merged correctly? Are sold units removed from active campaigns quickly? Are service histories tied to the right household? Are opt-ins documented?
- Process discipline: Do managers use the same lead statuses, appointment outcomes, recon stages, and lost reasons every day? Or does each department have its own dialect?
- System connectivity: Does the CRM know what happened in service? Does the service lane know equity position? Does inventory status update fast enough to avoid bad outreach?
- Manager accountability: Is someone reviewing exceptions weekly, or did the tool get installed and become “the vendor’s thing”?
Most dealers want to start at layer four with dashboards, AI summaries, predictive campaigns, and automated offers. The better stores start lower. They clean the inputs first because they know one ugly customer experience can cost more than a month of software savings.
Where the Breakage Shows Up First
The first place weak infrastructure exposes itself is customer communication. A wrong price in an email is embarrassing. A wrong payment quote in a text thread is worse. A service customer getting an equity message on a vehicle they traded six months ago makes your store look disorganized, not technologically advanced.
The second weak spot is inventory. If your vehicle status discipline is sloppy, AI-assisted merchandising and pricing get suspect fast. I’ve watched stores let automated campaigns promote units still in recon, units with missing photos, and units that had already been penciled on Saturday. Then the sales manager blames the tool. Sometimes the tool deserves it. Plenty of them are oversold. But a lot of the time, the store is feeding it garbage.
The third spot is the service lane. This one matters because service data is one of the few advantages franchised dealers still own outright. You know the RO history, mileage, declined work, ownership cycle, and relationship temperature better than any shared lead provider or auction lane. But if the advisor notes are thin, phone numbers are bad, equity mining rules are generic, and nobody owns the handoff to sales, AI just creates more half-qualified noise.
A Back-of-Napkin AI Readiness Score
If I were running a rooftop, I’d score five areas before approving another automation project. Give each one 0, 1, or 2 points.
| Area | 0 Points | 1 Point | 2 Points |
|---|---|---|---|
| Customer records | Duplicates everywhere | Cleaned when obvious | Weekly merge review |
| Inventory status | Often stale | Usually right by next day | Same-day status discipline |
| Lead outcomes | Rep-by-rep interpretation | Basic definitions | Manager-audited standards |
| Service-to-sales handoff | Advisor memory | Occasional spreadsheet | Documented workflow |
| Compliance and opt-in | Assumed | Checked when challenged | Tracked before outreach |
A store scoring 8 to 10 can probably handle more advanced automation without embarrassing itself. A store scoring 5 to 7 should automate narrow workflows first and watch them closely. Under 5, I’d spend 30 days cleaning process before buying anything with “AI” in the pitch deck.
That may sound conservative. Fine. I’ve been wrong before. I underestimated how quickly some AI writing tools would improve, and I thought more customers would resist automated texting than actually did. But I haven’t changed my mind on this: bad dealership process scales badly.
The Smart Use Case Is Narrow, Not Magical
The stores getting value aren’t asking AI to run the dealership. They’re using it to tighten specific workflows with clear rules. Confirm tomorrow’s service appointments. Identify service customers with equity and strong ownership signals. Send a timely text after a declined repair. Flag stale leads that need a manager touch. Summarize customer history before a call.
That’s where platforms like AutoRelay fit for dealers focused on customer communication and service-lane acquisition. Not as a magic layer that fixes the store, but as automation sitting on top of a workflow the operator already understands: find the right service customer, reach them through SMS, create a clean handoff, and measure acquired units against what auction or digital wholesale would have cost.
The math matters. If your auction-sourced unit costs you transport, buy fee, arbitration risk, heavier recon, and more days-to-frontline, then even a modest service-lane acquisition rate can change the month. But only if the customer data is clean enough to avoid wasting the opportunity.
Audit This Before You Add More AI
Pull 50 closed ROs from the last 30 days where the vehicle is between three and eight years old. Check four things: valid mobile number, text permission, current mileage, and whether the customer has an active sales or equity note. Then count how many records are actually usable for intelligent outreach.
If fewer than 35 of the 50 are usable, your next AI project is not an AI project. It’s a data and process cleanup project. Fix that first, then automate. See how AutoRelay helps dealers acquire inventory from their own service drive → getautorelay.com