AI Lead Follow-Up for Dealerships: Response Routing, Attribution and Proof

Turning on AI follow-up takes an afternoon. Knowing whether it's selling cars takes discipline. Most stores that are disappointed with AI texting weren't let down by the AI; they never decided what should happen to each reply, never tied sales back to sources, and never read what was being sent in their name. This is a practical guide to all three, whatever software you use.

1. Response routing: every reply has an owner

The first message is the easy part. The value is in what happens when the customer answers. Sort replies into a handful of routes and give each one an owner:

Reply	Example	Route
Ready to visit	"Can I see it Saturday?"	Book it, then a task for the rep
Question	"Does it have a tow package?"	AI answers from the vehicle record, or says it will check
Payment question	"What would I pay with $3,000 down?"	AI gives an estimate from your rates and fees, labeled as an estimate, and offers to have a manager confirm
Objection	"Found one cheaper in Nashua"	Manager or senior rep
Not now	"Maybe in the spring"	Long-term nurture with a date
Wrong number or STOP	"STOP"	End all contact, immediately, everywhere

Some vendors automate this directly; TECOBI, for example, describes a Response Bot that picks a route for each reply. However your system does it, write the routes down and check a sample every week. A ready buyer stuck in a nurture sequence is the most expensive mistake in follow-up.

2. Guardrails before volume

Before you let AI send anything on its own, set the guardrails:

• Draft first. Have the AI write and a person send for a week or two. Move to automatic only when you'd send the drafts as written.
• Quiet hours. 8 am to 9 pm is the federal telemarketing window for calls; it's a sensible default for texts too.
• A daily cap per customer. Three automated messages a day is plenty.
• Step back when a person steps in. If a rep just replied, the AI should hold off.
• STOP means stop. Never reply to an opt-out except to confirm it.
• Facts only from the record. Inventory, price and payment estimates should come from your data, not the AI's imagination.

3. Attribution: sales by source, not texts by source

Every lead needs a source that never gets overwritten: the website form, a listing site's ADF email, a phone call, a walk-in, a Facebook lead. Then measure each source on what matters:

• Minutes from lead to first reply.
• Leads to appointments set, appointments shown, and sales.
• Cost per sale: what the source cost this month divided by the cars it sold.

Be careful with credit. If the AI sent the first text but a rep had a 20-minute phone call and the customer came in from a Google search a week later, three things touched that sale. Pick a rule (first touch, last touch, or "any AI conversation in the 30 days before the sale") and use it consistently, so this month compares fairly with last month.

4. Proof: read it, count it, summarize it

The proof that AI follow-up works is not a dashboard of messages sent. It is:

1. Reading it. Pull ten AI conversations a week and read them like a customer would. Would you have replied?
2. Counting outcomes. Appointments, shows and sales per 100 leads, by source, against the same month last year or the month before you switched.
3. A daily summary you actually read. One short message each morning beats a report nobody opens.

A 30-day plan

1. Week 1: AI in draft mode. Write down your routes and owners. Record this month's baseline: response time, appointments, shows and sales by source.
2. Week 2: read every draft the team changed before sending. Fix the template or the instruction behind each change.
3. Week 3: automatic sending after hours only, with quiet hours, the daily cap and the human-reply hold on.
4. Week 4: compare against the baseline. Keep what moved appointments and shows; turn off what only moved message counts.

How CarSales AI OS does it

AI texting has three modes, set in Settings → AI Voice: off, draft (a ready-to-send reply waits in the Inbox) and automatic. In automatic it sends between 8 am and 9 pm, at most three messages per customer a day, holds off if a person replied in the last 30 minutes, and never answers STOP. It writes from your inventory and deal numbers and labels payments as estimates, in English or Spanish to match the customer. Ten follow-up plays (new lead, appointment, no-show, stalled deal, lost lead, be-back, dead deal, aged leads, new arrival or price drop, and sold) run on their own and stop when the customer replies.

Every lead keeps its source, and the Sales Scoreboard shows response time and leads to appointments, shows and sales by rep and by lead source. Every lead also gets a 0–100 score with the reason, and when a known customer opens a car link the rep gets a "looking at this car right now" alert. For the daily summary, turn on the morning brief and it texts you at 8:15 AM with new leads, today's appointments, overdue tasks and aging units.

If you're weighing an AI follow-up layer against an all-in-one system, our CarSales AI OS vs TECOBI comparison goes through it line by line. Pricing is per store with every feature included; get your price in 30 seconds.

Sources, checked October 2026

• TECOBI home page, product modules (tecobi.com)
• FCC: Telemarketing and robocalls

Vendor prices and features change. If anything here is out of date, email support@carsalessoftware.com and we will correct it.