Modern Dealership Leadership Means Managing AI at the Edge

At 7:42 a.m., a customer arrives for a loaner your store never reserved. The automated conversation says one was promised. The advisor says nobody approved it. The service manager is now choosing between eating a rental, disappointing the customer, or arguing over the wording of a text message before the first RO is closed.

That is modern dealership leadership in one ugly little scene. AI can handle more conversations, surface opportunities, summarize calls, prioritize leads and remove repetitive work. It can also create customer commitments faster than a manager can catch them. CBT News recently framed AI adoption in auto retail as a leadership and culture issue, and I think that is the useful way to look at it. The hard part is no longer deciding whether employees may use AI. They already are.

AI Moves Decisions Downstream

Most dealership operating structures were built around controlled access to information. Managers held the appraisal pencil, desked the deal, approved the goodwill adjustment and decided which customer received an exception. AI changes that flow. A salesperson can generate a follow-up campaign in seconds. An advisor can summarize service history without reading every line. A BDC agent can send polished answers on topics they do not fully understand.

That speed is valuable right up to the point where polish gets mistaken for authority. Customers do not care whether a promise came from a manager, an employee using an assistant, or an automated workflow. If it came from the dealership, they treat it as the dealership's word.

I've seen stores respond by locking everything down. That usually lasts until the team starts using personal tools outside the approved process. I've also seen stores turn employees loose with vague instructions to “use AI responsibly.” That is not a policy. It is an invitation to discover the guardrails through customer complaints.

The Manager's Job Becomes Exception Design

I used to think good automation mainly depended on choosing the right workflow. I was wrong. Workflow design matters, but exception design matters more. The routine customer gets handled quickly almost anywhere. Leadership shows up when the trade has an open recall, the quoted payment excludes an add-on, the service customer wants a diagnosis by text, or the system identifies equity that disappears after a repair bill lands.

A manager needs to define which situations stop the automation, who receives them and how quickly that person must act. Otherwise, AI simply creates a cleaner-looking version of the same old accountability problem.

Use the Tolerance Sheet

I'd argue that every customer-facing AI workflow needs a one-page tolerance sheet. Not a 40-page policy from legal. One page that the department manager owns and updates when the store learns something.

• Task: State exactly what the system may do. “Schedule an appraisal appointment” is clear. “Handle acquisition leads” is not.
• Tolerance: Define what it may quote or commit without approval, including appointment windows, incentives, trade ranges, rentals and service discounts.
• Tripwire: List the conditions that require a human handoff, such as negative equity, title issues, safety concerns, repeated objections or an upset customer.
• Owner: Name one manager accountable for reviewing exceptions and changing the workflow. A committee cannot own a missed appointment.

The tolerance sheet also exposes a culture problem many stores would rather ignore: managers sometimes disagree on what the process actually is. AI did not create that inconsistency. It just makes the inconsistency operate at scale.

Do Not Measure AI by Activity

Message volume, response speed and tasks completed look good in a vendor recap, but they are weak management measures by themselves. A system can send 4,000 texts and still produce no additional sold units, acquired vehicles or retained ROs. Worse, it can improve response time while increasing opt-outs and customer frustration.

Measure the business result and the failure rate together. For a sales workflow, pair appointments shown with complaint or opt-out rate. In service, pair declined-work recovery with discount leakage. For acquisition, pair appraisal appointments with vehicles bought, acquisition cost and days-to-front-line. Speed without quality is just a faster way to make a mess.

Audit Five Conversations, Not Fifty Dashboards

Pull five AI-assisted customer conversations from each department every week: one conversion, one loss, one opt-out, one escalation and one selected at random. Have the department head answer four questions:

1. Did the customer receive a clear and accurate answer?
2. Did the workflow create a commitment the store could not reliably keep?
3. Was the handoff early enough for a manager to change the outcome?
4. What rule should be adjusted before the same situation repeats?

Then pull the associated outcome from the CRM or DMS. If managers cannot connect the conversation to an appointment, RO, appraisal, sold unit, opt-out or concession, they are reviewing technology theater instead of performance. Modern leadership is not being enthusiastic about AI. It is deciding where the machine may act, where a person must step in and who owns the result when the line gets crossed.