AI Lead Qualification for Car Dealerships

Qualification is what you learn in the first conversation, not a number a model stamps on a lead before anyone speaks to the customer.

Ask three vendors what AI lead qualification means and you will get two categories of product described with the same words.

The first is prediction. A model looks at a lead's attributes and produces a ranking that tells your team who to call first. It is appealing because it promises to solve the oldest problem in a BDC, which is that there are more leads than hours.

The second is conversation. Software handles the opening exchange, asks the questions a human would ask, records what the customer said, and hands a salesperson a lead they already know something about. This does not rank anyone. It makes the first contact happen faster and captures the answer.

These are not variations on a theme. One replaces judgment, the other replaces typing and waiting. Most of the disappointment we hear about in this category comes from stores that bought the first and needed the second. Our position, stated plainly below, is that the second is where the reliable money is.

This will read as a strange thing to put on a page about qualification, so let us be direct about it.

Nothing we deliver is filtered, scored, ranked or quality checked. We ask pre screening questions at the point of capture, and every submitted lead inside your territory is delivered to you exclusively. If a lead has a genuine problem, that is handled by post delivery replacement review rather than by a promise made up front about quality that no vendor can actually keep.

There is a reason for that beyond modesty. A ranking model is only as good as the outcome data it learned from, and dealership outcome data is contaminated by exactly the behavior the model is supposed to fix. If your team historically ignored leads with no phone number, the model learns that those leads do not close, and it recommends ignoring them. The prediction is right and useless at the same time. Meanwhile the customer who typed a bad number and answered every email is sitting at the bottom of the list.

The honest version is harder and more effective: contact everyone quickly, learn what is actually true from the conversation, and let your people apply judgment to real information. Our page on measuring lead quality covers how to evaluate a source without pretending you can grade an individual lead in advance.

The most reliable qualification signal is not inferred. It is asked.

Lead pages give you a landing page builder with per page URL settings, so the form a shopper fills in can ask what you actually need to know for that campaign. A trade acquisition campaign should ask about the vehicle, mileage and payoff situation. A finance focused campaign should ask about the timeline and the down payment, not about trim preference. A model specific campaign should ask what they are driving now.

Two cautions, and both matter. Every additional field costs you completions, so ask for the two or three answers that genuinely change how you would work the lead and drop the rest. And for anything financing related, fair lending rules forbid narrowing an audience by age, gender, income, marital status, household size, education, language or ZIP. A vendor offering to help you target around those rules is handing you liability and calling it a feature.

What comes back is genuinely useful and it did not require a model. It required asking.

The AI capability we would actually defend in front of a skeptical GM is call transcription, and it changes management more than it changes selling.

Every call through the platform can be recorded and transcribed. A manager reads a twelve minute conversation in thirty seconds instead of listening to it. That single change makes coaching possible at volume. You can see which salesperson never asks about a trade, which one talks past the customer's budget question, and which one is actually good and deserves the next up. You can settle a customer dispute by reading what was said rather than reconstructing it from two memories.

It also solves the qualification handoff. A BDC agent takes a call, and the salesperson picking up the customer reads what was discussed instead of asking the same five questions again. Customers notice being asked twice, and they read it as a store that does not talk to itself.

Document AI handles the other half of the capture problem: VIN scanning from a phone camera and document reading, so a trade or an identification document gets into the record accurately without anyone typing. More on the call transcription page.

If you only fix one thing, fix time to first response, because it is the most controllable variable you have and it is almost always worse than managers believe.

Automations and follow up processes let you build the first touch once and let it run: an immediate text acknowledging the inquiry, an email with the specific vehicle, a task assigned to a named person with a due time. RCS with automatic SMS fallback means the message looks like it came from a business rather than a stranger, without cutting out anyone on an older handset.

After hours is where most stores quietly lose. A lead arriving at 9:40pm gets worked at 10am, by which point the shopper has heard from two competitors. An automated first response does not close anybody, and we would not claim it does, but it holds the conversation open until a person is available. Set the expectation honestly in the message rather than pretending a human is typing.

Inbound calls have the same problem in a different shape. AutoMail provides IVR, call routing, call forwarding and number management, so a call at 6pm reaches somebody instead of ringing into an empty tower. Missed calls become a follow up task rather than a lost customer. See after hours lead response and response time and close rate.

Here is the uncomfortable observation. When a store says it needs lead scoring, what it usually needs is for the right lead to reach the right person immediately. That is a routing problem with a known answer, not a prediction problem.

Global lead distribution with rules, layered over per store settings and distribution lists, covers the cases stores actually care about. Send everything from a specific campaign to the internet director. Round robin the general inquiries across the floor with escalation if nobody touches them inside a set window. Route a Spanish language inquiry to the two people who speak Spanish. Send acquisition leads to whoever runs the buying desk. Keep a global lead history so that when a general manager insists their store never received something, you can show where it went and when.

None of that requires a model, all of it is auditable, and unlike a score, it is something you can explain to the person it affects. The lead routing rules page goes deeper, and round robin distribution covers the floor case.

A lead is not qualified or unqualified. It is a person whose timing you do not know yet. That reframing changes what software should be doing.

Most stores stop after two or three touches, and most managers believe their team does more than that. Pull the actual numbers before you buy anything, because the gap between the two is usually where the missing gross is. A shopper who was three months out in March is buying in June, and the store still in the conversation gets the appointment.

Drip leads with a drip editor, follow up processes, task automation and reminders exist so a twelve touch cadence is no harder to run than a three touch one. Appointments and calendar tie the cadence to something real. Unified messaging means the thread survives across text, email and voice instead of resetting each time.

The discipline point that no software fixes: a cadence only works if opt outs are honored and the messages are worth reading. Consent is handled across the platform rather than per department, and an unsubscribe is information, not an obstacle. See follow up cadence.

Take these into the demo and watch how comfortable the vendor is being pinned down.

1. What data was the model trained on, and is any of it from my store? A model trained on other dealers' outcomes is describing their sales process, not yours.
2. What happens to a lead the model rates low? If the answer is that it goes to the bottom of a list nobody reaches, you have automated neglect.
3. Can I see a transcript or recording of what the AI said to a customer? If not, you are accountable for messages you cannot review.
4. Does the customer know they are talking to software? Decide your own policy on this before a customer decides it for you in a review.
5. How are opt outs handled across every channel? One consent record, or four, is the whole compliance question.
6. What is the measured improvement, on what baseline? A vendor quoting a lift number before seeing your data is guessing at your expense.

The AI vendor evaluation guide expands each of these, and predictive lead scoring covers the prediction side on its own terms.

To close without ambiguity. Provided: pre screening questions at capture through lead pages, instant automated first response, follow up processes and drip cadences, task automation and reminders, global lead distribution with rules and distribution lists, SMS and MMS, RCS with SMS fallback, a VoIP dialer with call recording and transcription, voicemail drop, IVR and call routing through AutoMail, email with a real inbox and composer, an email validator and a phone validator, document AI for VIN scanning and document reading, appointments, and three layers of reporting.

Not provided: lead scoring, ranking or grading of any kind. No quality tier, no filtering, no promise that the leads you receive have been screened for you beyond the questions asked at capture. We also do not do lender submission or automated credit decisioning.

Pricing is month to month with no long term contract. CRM Only starts at $199 a month if you bring your own lead flow. Programs including exclusive local leads start at $799, and combined buyer and seller programs at $1,599. Detail on the pricing page. We cannot guarantee close rates or appointment volume, because those depend on your team and your market. Tell us what your follow up looks like today and we will tell you honestly whether this helps.

Does LeadLocate score or rank the leads it delivers?

No. Nothing is filtered, scored or graded. We ask pre screening questions at capture and deliver every submitted lead inside your territory exclusively. Genuine problems are handled by post delivery replacement review.

Then what is the AI actually doing?

Practical work: transcribing and summarizing calls so managers can coach at volume, reading documents and VINs from a phone camera, and driving automated first response and follow up cadences so leads are contacted quickly and consistently.

Can it answer leads automatically after hours?

Yes, through automations and follow up processes, and inbound calls are handled by IVR and routing. An automated first touch holds the conversation open until a person is available. We would not claim it closes anyone.

Is predictive lead scoring worth buying from someone else?

It can be, if you have enough clean outcome data of your own and you use the output to prioritize rather than to ignore. Ask what the model was trained on and what happens to leads it rates low before you commit.

Can we ask qualifying questions on our own forms?

Yes. Lead pages let you build the form and ask what matters for that campaign. Keep it to the two or three answers that change how you work the lead, because every extra field costs completions.

Are there rules about what we can ask in financing campaigns?

Yes. Fair lending rules forbid narrowing a financing audience by age, gender, income, marital status, household size, education, language or ZIP. Any vendor offering to work around that is offering you liability.

We will show you a call transcribed, a cadence running and routing rules sending the right lead to the right person. Month to month, no long term contract.