The automotive industry doesn’t have an AI adoption problem anymore. It has an AI control problem.
Dealers have spent the last few years being sold a simple promise: plug in AI, watch leads convert, go home early. And to be fair, the technology has delivered real results. But something interesting is happening now that the early hype has cooled and dealers are actually living with these systems day-to-day.
They’re not asking for more automation. They’re asking for more control.
We analyzed how dealerships across the US customize their AI agents when given real control over behavior and here is what we found:
The evidence is hiding in the instructions
When you give dealers the ability to customize how an AI agent behaves — what it says, when it says it, how hard it pushes, and when it stops — you’d expect most of them to leave the defaults alone. That’s how most software works. Set it and forget it.
That’s not what’s happening.
Across thousands of dealerships actively modifying their AI agent behavior, a clear pattern has emerged. Dealers aren’t tweaking. They’re authoring. They’re writing new customer journeys from scratch, encoding sales strategy directly into the AI’s instructions, enforcing compliance boundaries, and calibrating tone to match their unique brand and voice.
This isn’t configuration. It’s a strategic operating layer. And the patterns reveal something the industry should pay close attention to.
What the best dealers actually want from AI
Forget the feature wars for a second. When dealers get real control, here’s what they consistently optimize for:
Driving a specific outcome, not a generic one. The top-performing dealers aren’t telling AI to “engage the customer.” They’re telling it to schedule an in-person visit. Or set up a call with a commercial rep. Or confirm availability — and nothing more. Every journey has a defined destination. The AI isn’t wandering. It’s executing.
Qualifying faster, not harder. Dealers are embedding trade-in questions and discovery prompts directly into the AI’s first response, but only when those questions increase deal value or resolve ambiguity blocking the next step. They’re not running full needs assessments through a chatbot. They’re capturing high-signal data points early and when contextually relevant, then moving.
Protecting the deal before advancing it. Across the board, dealers are drawing hard lines around what AI can and cannot say about pricing, payments, and inventory availability. Not because they’re being cautious, but because they’ve learned that premature financial specificity creates downstream friction. Once the AI mentions a payment figure, the customer treats it as a promise. Even when it’s caveated. Smart dealers have figured out that helpful ambiguity beats precise inaccuracy every time.
Matching urgency to intent. This might be the most sophisticated pattern in the data. Dealers are rejecting the assumption that every lead wants to buy right now. They’re calibrating tone and urgency based on where the customer actually is, not where the dealership wishes they were. Luxury brands get less pressure, not more. Exploratory shoppers get guidance, not a hard close. High-intent buyers get urgency and specific appointment times. The AI adapts–or at least, it should.
Knowing when to stop. Here’s the signal that separates the best dealers from the rest: they’re defining explicit rules for when the AI should hand off to a human. Not as a failure condition — as a success condition. Financing questions? Escalate. Complex trade scenarios? Escalate. The AI knowing its lane is what gives dealers the confidence to let it operate autonomously everywhere else.
The real story isn’t about AI. It’s about strategy.
This data reveals how the best dealers think.
They’re not treating AI as a replacement for process. They’re treating it as an amplifier of process. The dealers who are getting the most out of their AI are the ones who had a clear sales strategy (supported by results) before the AI showed up and are now encoding that strategy into every automated touchpoint.
The AI becomes the execution layer. The dealer remains the strategist.
That’s a fundamentally different model than what most of the industry has been sold. The pitch has been: “Let AI handle it.” The reality is: the dealers who let AI handle it without encoding their own playbook and embedding it systemically across their workflows are the ones getting generic results.
Five undeniable principles from the data
For dealer principals and GMs thinking about how to get more from AI (or evaluating what to invest in next), here’s what thousands of dealerships are telling us through their actions:
Start with the goal, not the message. Every customer journey should have a defined outcome. If the AI doesn’t know what success looks like, it’s just generating words.
Treat constraints as features. “Do not discuss pricing” isn’t a limitation. It’s a trust builder. More often than not, constraints are purposely defined and designed. The guardrails are what make autonomous operation possible.
Ask fewer, better questions. Discovery should capture signals that move the deal forward. If a question doesn’t increase deal value or resolve an obstacle, cut it.
Campaign awareness is table stakes. The AI’s first response should reflect what the store is actually selling right now, and the lead source the inquiry came from. If it’s March and the AI is still running January’s playbook, you’re leaving money on the table.
Make the handoff a strategy, not a safety net. Define when and why the AI steps aside. Make it intentional. Make it clean. And make it seamless: the customer should feel like a concierge team is in sync and working for them. A good AI knows when not to answer and how to bring in humans artfully. The best dealers are designing for exactly that.
Where this is headed
The dealerships that figure this out first won’t just run more efficient operations. They’ll build something harder to quantify and even harder to replicate: an operating rhythm supported by an AI operating system that reflects their brand, encodes their strategy, and compounds in value over time.
The ones that don’t will keep running somebody else’s default playbook — and wondering why the results feel generic.
The future of AI in automotive retail isn’t about who has the smartest model. It’s about who programs it with the smartest strategy.
And right now, the best dealers aren’t waiting to be told how. They’re already writing the playbook.
Methodology
This analysis is based on a proprietary dataset of AI agent configuration instructions authored by thousands of Impel customers as of December 31, 2025. The dataset captures dealer-written modifications across multiple customer journey types — including outreach, vehicle interest, digital retailing, financing, trade-in, and service flows. Patterns were identified by categorizing instruction themes across all modified journeys, with particular attention to recurring strategic behaviors: goal-setting, qualification logic, financial guardrails, tone calibration, campaign integration, and human handoff rules. No customer data or personally identifiable information was included in the analysis.