Picture your assistant service manager at 4:45 p.m., reopening an AI-generated conversation because the customer says somebody promised an update. The advisor never saw it. The BDC thought service owned it. Everyone completed the training. Now a manager who should be clearing the drive is reconstructing a conversation your new system was supposed to make easier.
That isn't necessarily an AI accuracy problem; it's an ownership problem with a software invoice attached.
I'd argue that employee involvement matters most after the demonstration, when somebody has to decide what happens to an unfinished customer request. Knowing how to use a tool is different from trusting the work it hands you. Before asking whether the team has embraced AI, ask whether employees can accept that work without hunting for context, negotiating responsibility or making a promise they cannot keep.
Dealership AI adoption starts with workable ownership
A dealer principal sees faster responses. An advisor sees another place to check before telling a customer what's happening. A GSM sees more opportunities, while a salesperson sees appointments that may arrive without enough context. A Used Car Manager may see a chance to buy inventory—but only if someone can move an interested customer toward an appraisal while that customer is still available.
These aren't necessarily competing attitudes toward technology. They're different exposures to the same work. The person approving the subscription rarely absorbs every exception it creates, and the person handling those exceptions may have had no say in the setup. Employees have good reason to judge a change by what lands on their desk rather than what happens in a demonstration.
Compensation can widen that gap. Ask an advisor to interrupt a repair-order conversation to help acquire a vehicle, and you've assigned work with an uncertain payoff. Ask a salesperson to monitor additional requests without clarifying opportunity credit, and don't act surprised when established follow-up remains the priority. Set expectations for participation, recognition and credit before judging enthusiasm. Calling someone resistant won't resolve a conflict between the work you request and the work you reward.
Logins measure access, not adoption
Training completion tells you who attended. Login counts tell you who opened the application. Neither tells you whether employees trust the process enough to stop maintaining a parallel one. For a department manager, the more useful question is whether a customer request reaches someone who can act without another employee having to shepherd it there.
Look for the shadow work: screenshots sent to managers, handwritten callback lists, duplicate CRM notes, somebody walking across the showroom to confirm an assignment. Some redundancy is sensible during a rollout. When it becomes permanent, the store is paying for automation and paying employees to insure against it.
Here's a hypothetical audit of 40 consecutive AI-to-employee handoffs. Eleven require someone to chase down an owner; the other 29 are individually checked and confirmed to have a named, available owner and a clear next action. If each ownership chase consumes five combined staff minutes, that's 55 minutes spent finding someone to act—not helping the customer. This doesn't prove the tool loses money, but it identifies labor that a response-time report may miss.
A named owner who is off that day doesn't count as coverage.
Give employees authority over the exceptions
Start with one workflow and include the people who receive its unfinished business. For service-lane acquisition opportunities, that means hearing from the advisor, the employee responsible for customer follow-up and the Used Car Manager—not simply asking department heads to approve a plan. Have them settle four practical questions before expanding it:
- The boundary: which requests can proceed without employee review, and which commitments require an authorized person. Customers should not have to untangle a promise the store cannot honor.
- The owner: who accepts each handoff, with a named backup during lunch, days off and closing. Coverage should match the response time the customer is actually given.
- The recovery: who can stop further promises, correct a customer-facing mistake or seek management approval. Document when escalation is required and who covers an unavailable manager.
- The credit: how work and opportunities are recorded so employees aren't penalized for participating. Make acquisition credit and follow-up expectations clear across service and used vehicles.
Consider a hypothetical service customer who asks about selling her vehicle while waiting for repairs, but the person authorized to appraise it is unavailable. The advisor should not have to abandon the drive or invent a purchase figure to preserve the opportunity. A designated acquisition contact can own the next customer conversation, confirm whether an authorized backup can inspect the vehicle and, if necessary, agree on a specific follow-up time. The Used Car Manager retains the valuation decision; the contact owns keeping the customer informed. That distinction gives the store a better chance to acquire inventory without leaving service responsible for an offer it cannot approve.
Train against those exceptions, not just the clean demonstration. A customer changes the subject, an assigned employee leaves early or a message suggests an unapproved commitment. Asking a manager can be the correct next step when the escalation path and expected timing are documented. The warning sign is an improvised search for permission while the customer waits. Review whether the promised follow-up happened, not merely whether somebody accepted the request.
At your next operating review, examine 40 consecutive handoffs from one workflow, verify ownership, availability and next action, and classify misses as ownership, coverage, authority or missing information. Fix the largest category with the employees doing the work, then repeat the review to see whether fewer customers—and coworkers—are left chasing an answer.