Everybody’s talking about AI. It’s become one of those words that gets thrown around so much it almost loses meaning. And if you’re trying to run a dealership, it can feel overwhelming fast.

AI for fixed ops is no exception. Before you spend time, money or your team’s patience on something new, here’s a grounded, six-step process you can apply to the solutions you’re considering.

1. Start With the Use Case.

The biggest advice I give dealers is to walk into the conversation with a crystal-clear idea of what you’re trying to accomplish.

If you’re in pure discovery mode, that’s fine. Learn. Ask questions. But the minute you’re evaluating vendors, you should be able to say, “Here’s the workflow that’s breaking, here’s the metric we need to move, and here’s what success looks like.”

What KPI are you actually trying to move? Appointment set rate? Call answer rate? Advisor time per RO?

In service, I usually see two common starting points:

• “We can’t handle the call volume.” There are just too many calls. All your advisors are super busy. They’re not able to service your customers. Your goal is coverage.
• “We can handle calls, but we’re burning good people on repetitive stuff.” Maybe you’re staffed well. You’ve got a BDC that’s taking tons of calls, answering questions about hours, oil changes, recalls, tow-ins. Now the goal isn’t coverage; it’s redeployment. How do you free those people up for more valuable work?

Those are two very different situations. And if you don’t define which one you’re in, you’ll end up evaluating solutions the wrong way.

2. Do Some Prep Work.

This category is moving fast. These language models are improving constantly.

Because of that, I really encourage dealers to do some prep before diving in. You don’t need to get technical. The more baseline knowledge you bring into a conversation with a vendor, the more you’re going to get out of it and the easier it is to set clear expectations.

If you want a starting point, Andrew Ng’s “AI for Everyone” at DeepLearning.ai is a great beginning course. It’s not too deep. It just gives you enough foundation to understand the landscape.

3. Is It Real AI or a Dressed-Up Phone Tree?

A lot of dealers think they’re buying AI when they’re really buying a slightly upgraded interactive voice response system. Traditional IVRs are rigid: Press one for this, press two for that. They have pre-canned responses and only react to specific keywords they recognize.

A true voice assistant listens to what the customer is saying and responds in context. If someone calls and says, “Hey, it’s my birthday, and I need to change my flight,” a basic system just hears “change flight.” A smarter system might say, “Happy birthday. I can help you change that.”

It sounds small, but customers can tell pretty quickly if they’re talking to something intelligent or just another phone tree. If it feels like a phone tree, they’ll press zero or ask for a person almost immediately.

4. Evaluate Using the Art of AI.

So when we talk to dealers about how to evaluate these tools, we usually come back to a three-part framework called the “art of AI”:

• Adoption: Are customers actually going to speak to it? If they don’t adopt it, nothing else matters. Small things help here like voice quality, tone and tailoring accents or language. It has to feel natural enough that customers stay on the line.
• Resolution: Once they’re talking to it, can it solve the reason they called? What’s the resolution rate? How many calls get handled without a transfer? Can it book the appointment or answer the question without bouncing the call back to your team?
• Tailoring: This is the big one. Honestly, it’s my favorite part of the framework, because it’s the most critical. Does it behave the way you run your store? If a customer pushes back on a diagnostic fee, what should happen? Should it ask again? Should it transfer? Should it hold the line? You already train your advisors for these moments. You want to train AI the same way.

5. Pick Your Spot.

Another thing to watch is placement in the call flow. If it answers every call, 24/7, it’s going to get everything: sales calls, wrong numbers, oddball questions, angry customers, everything.

If it sits behind “press two for service,” the calls will be more focused, and the metrics will look different.

Also remember that you don’t have to go all-in on day one. You can start lower in the tree as a safety net, then move it up as you get comfortable.

6. Don’t Lose the Human Touch.

There are still conversations that your team should handle. Status calls. Difficult updates. Customers who are upset or irate. Those moments still need the human touch, and good systems should recognize that quickly and offer a live transfer.

One last piece of advice: Go at your speed. The “start now or get left behind” message gets thrown around a lot, but it’s not helpful.

Know what you’re trying to accomplish and get clear about the use case. Define what winning looks like in your store. Then ask vendors to show you real examples of dealers in situations like yours and what results they actually achieved.

Sean Hartman is vice president of sales for Toma.