A dealership can answer a web lead in minutes by sending every enquiry into one place, letting an AI step draft a reply from the lead's own words and your stock data, and asking a salesperson to approve it with one click. Build it in Zapier or Make for roughly $10 to $60 a month.
The catch is that the AI never decides facts. Price, availability, finance and trade-in values come from your systems or a person; the AI only writes the sentences around them. Keep that boundary and the speed gain is real, because most of the delay in a first reply is the time it takes somebody to open the lead, find the car and type.
Where the minutes go in a typical first reply
Before building anything, time one real lead by hand. A buyer fills in a form on a stock page at 6.40pm. The email lands in a shared inbox, somebody notices it after the showroom closes or after a test drive finishes, searches the stock list for the car, checks it is still unsold, writes three paragraphs and sends. None of those steps is hard. Together they take a quarter of an hour on a good day and the following morning on a bad one.
An automation removes the noticing, the searching and the typing. It leaves the judgement with your staff. That is why the design below ends with a human approval rather than an automatic send: the lead is answered in minutes because a draft is waiting, not because nobody looked at it.
The chain in five links
| Link | What happens | Who or what does it |
|---|---|---|
| 1. Capture | The lead arrives from your website form, a marketplace listing email or a chat widget | A webhook, or a mailbox trigger |
| 2. Record | A contact and an enquiry are created or updated in the CRM, with the vehicle reference | The CRM |
| 3. Look up | The advertised price and sold status for that vehicle are fetched from your stock sheet or feed | A lookup step |
| 4. Draft | An AI step writes a short reply using only the fields supplied | AI by Zapier, or an AI module in Make |
| 5. Approve | The draft lands in the salesperson's email drafts and a chat message pings them | A person |
Everything else in this tutorial is detail on those five links. If your website chatbot already captures leads, it simply becomes another source for link 1.
Tidy the inputs before you automate anything
Automations amplify whatever they are fed. Three jobs come first and take about an afternoon.
- List every place a lead can arrive. Your own site form, marketplace listing emails, a social ad form, a phone call logged by hand. Most marketplaces forward leads as emails with a fairly fixed layout; check yours by looking at five recent ones.
- Add a vehicle reference to every form. A hidden field holding the stock number or registration means the lookup in link 3 is exact. Without it the AI has to guess which car "the blue one" means.
- Put prices and sold status in one place. A shared sheet with columns for stock number, make, model, year, advertised price, status and mileage is enough. If your dealer management system can export or sync that, use it; if not, a sheet updated each morning works for a small lot.
Here is the kind of messy forwarded lead the flow has to handle, and the clean fields you want out of it.
RAW EMAIL (illustrative)
Subject: New enquiry - Stock #4471
Name: [first name, typed in lower case]
Phone: 555-0142
Message: is this still avail? can i come sat, part ex my old hatchback
CLEAN FIELDS
first_name: [First name, capitalised]
stock_number: 4471
contact: phone and email as supplied
wants: availability, Saturday visit, part-exchange
For a form you control, ask for these fields directly and skip the parsing. For forwarded emails, an AI extraction step can pull the fields, and a filter can stop anything that is not a lead (newsletters, receipts) from reaching the draft step.
Building the flow in Zapier
Zapier counts a task for each successful action step; triggers, filters and Formatter steps do not count. An "AI by Zapier" step costs 1, 3 or 5 tasks a run depending on the model tier, or 1 if you bring your own API key. Here is a build that fits one Zap.
- Trigger. "Webhooks by Zapier" for a site form (Professional plan or above; it is not on Free), or a Gmail trigger on a label such as "Leads" for forwarded emails. Webhooks respond instantly; Zapier's Professional plan checks mailbox triggers about every two minutes.
- Filter. Continue only if the stock number is not empty and the message is not an auto-reply. This step is free.
- Find or create the CRM contact. HubSpot has a free CRM with a find-or-create contact action. This is one task. Save the enquiry as a note or deal at the same time.
- Look up the vehicle. A Google Sheets "lookup spreadsheet row" step matching on stock number. One task. Stop the Zap with a filter if the status says Sold, and route to a different draft (see the table of lead types below).
- AI step. Pass first name, message, vehicle fields and your rules. Ask for JSON with a subject and body so the next step can use them.
- Create a Gmail draft. Addressed to the lead, from the assigned salesperson's mailbox. One task.
- Notify. A Slack or Teams message to the sales channel: "Draft ready for [first name], stock 4471, 6.42pm" with the link. One task.
That is roughly five to nine tasks per lead depending on the AI tier. A lot with 200 web leads a month would use somewhere between 1,000 and 1,800 tasks, which is above the 750 included in the entry Professional plan ($19.99 a month billed annually, $29.99 monthly). Check the plan picker for the next tier up, and decide on your real lead count, not on a guess.
Zapier's Human in the Loop approval step is tempting here, but on the Professional plan its approvals can only go to yourself, so the draft-in-the-inbox method above suits a sales team better.
The same flow in Make
Make runs the same chain as a scenario: a Custom Webhook, a HubSpot module, a Google Sheets search, an AI module, a Gmail "create a draft" module and a Slack message. Make is credit-based, entry plans start at about $9 a month for 5,000 credits, and each module run uses a credit. Seven modules for 200 leads is about 1,400 credits. The AI step costs extra on top: Make's own AI provider can use more than one credit per run, and if you connect your own model key the tokens are billed by that provider instead. Routers use no credits, which is handy for splitting leads by type, and a bundle stopped by a filter uses none.
Use an instant webhook as the trigger. Make's scheduled triggers use a credit on every check even when nothing arrives, so a 15-minute schedule burns about 2,880 credits a month doing nothing. Make's own Human in the Loop app is Enterprise-only, so again use the Gmail draft as the approval step.
Pick by who will maintain it. If a person at the dealership will tweak the flow monthly, Zapier's step-by-step editor is gentler. If you have a part-time admin comfortable with branching diagrams, Make's routers make the lead-type split tidier and the credits go further.
The prompt, with sample output
Put the rules in the prompt, and pass the facts as fields. This is a prompt block for the AI step.
You write a first reply to a car dealership's web enquiry. Use ONLY the facts below.
If a fact is missing, say a team member will confirm it - never guess.
Rules:
- Max 110 words, plain and friendly, no exclamation marks.
- Repeat the advertised price exactly as given. Never calculate payments, discounts,
finance rates or trade-in values.
- If status is Sold, say so, name one similar vehicle only if "alternative" is filled in.
- If the lead mentions a part-exchange, say a person will value it in person.
- Offer two visit slots from "slots". Do not promise a test drive time.
- Sign off with the salesperson's first name.
Facts:
lead_first_name: {{first_name}}
lead_message: {{message}}
vehicle: {{year}} {{make}} {{model}}, {{mileage}} miles
advertised_price: {{price}}
status: {{status}}
alternative: {{alternative}}
slots: {{slots}}
salesperson: {{salesperson}}
An illustrative output for that enquiry:
Subject: Your enquiry about the 2019 hatchback, stock 4471
Hi [first name],
Thanks for getting in touch. The 2019 hatchback (41,200 miles) is still available at
the advertised price of $12,450.
We would be happy to see you on Saturday. We have 10.30am or 1.15pm free. If you bring
your current hatchback, one of the team will look it over and give you a valuation in
person, as we cannot do that by email.
Reply with the slot you prefer and I will have the car ready.
Kind regards,
[salesperson's first name]
What you would fix: the draft promises "have the car ready", which commits a person to a task. Add "do not promise preparation" to the rules if your staff dislike it. Also check the mileage and price against the sheet on every draft in the first weeks; those two fields are where a wrong lookup would show.
Different leads need different drafts
One prompt does not fit every enquiry. Route on lead type early, using a Paths step in Zapier or a router in Make.
| Lead type | Draft should | Must not |
|---|---|---|
| Car available | Confirm availability, repeat the price, offer two slots | Offer a discount |
| Car sold | Say it has gone, offer one alternative or a stock alert | Pretend it is still there |
| Part-exchange mentioned | Ask for make, year, mileage and photos; explain a person values it | Name a value or a range |
| Finance question | Say an adviser will call, ask for a preferred time | Quote rates, approve anyone, or promise acceptance |
| Out of hours | Auto-acknowledge with the real next-reply time | Promise a callback you cannot keep |
| Service or parts | Forward to the service desk with the vehicle noted | Treat it as a sales lead |
Finance deserves particular care. Credit rules differ between markets and a casual line such as "you'll easily get approved" can create real trouble. Have the draft hand finance questions to a person, and ask your compliance adviser what your own wording must include.
A six-person used-car lot, with numbers
Picture a six-person independent lot with 35 cars on the forecourt and about 160 web leads a month, three quarters of them arriving by marketplace email and the rest by its own site. Before the automation, the median first reply (measured by pulling timestamps from the inbox for 30 leads) was 3 hours 10 minutes, and after 6pm leads waited until the next morning. These figures are illustrative, so replace them with your own timings.
The owner builds the Make version over two evenings: webhook for the site form, a Gmail trigger for marketplace emails, a router with four routes (available, sold, part-exchange, finance), HubSpot's free CRM and a shared stock sheet. Salespeople get a Slack ping and approve from their phone. After a month the median first reply is 12 minutes during opening hours, and out-of-hours leads receive an acknowledgement within a minute with a real "by 9am" promise that the morning shift keeps.
The cost line looks like this for that lot:
| Item | Monthly (illustrative) |
|---|---|
| Make entry plan (about 1,300 credits of 5,000) | about $9 |
| AI model calls, 160 drafts at roughly 1,500 tokens in and 400 out on a mid-priced model ($2 / $10 per million tokens) | about $1 |
| HubSpot free CRM | $0 |
| Shared sheet and Gmail drafts | $0 on existing accounts |
| Salesperson time: about 40 seconds to review each draft, 160 leads | under two hours |
On Zapier the same flow would sit nearer $20 to $60 depending on the tier your task count needs. The software cost is small next to one extra car sold, but do not claim any particular number of extra sales; measure them.
Agreeing the approval rule with your sales team
Automation fails at dealerships for a human reason more often than a technical one: the draft is ready and nobody presses send. Agree three small rules before launch and write them where the team will see them.
- A named owner per lead. The CRM assigns each lead to one salesperson by rota or by vehicle. The Slack ping names that person, not the channel. A message addressed to everybody is read by nobody.
- A ten-minute nudge. If the draft has not been sent after ten minutes in opening hours, a second step pings the sales manager. In Zapier that is a delay step followed by a check of the draft folder; in Make it is a sleep module and a second search. Either costs a task or two per lead, so decide whether to run it on every lead or only on leads that mention a visit or a price.
- Edit freely, but edit the fact first. If the salesperson changes a price or slot in the draft, they also fix the sheet or the CRM, so the next lead for that car is right.
Say a sold-car lead goes wrong. The sheet still said Available, so the draft read: "The 2018 estate is available at $9,900." The salesperson knew it had sold that morning, rewrote the reply to offer a similar car, and marked the sheet Sold. The error was caught only because a person read the draft, which is the whole argument for approval in month one. The fix in the automation was to add a "sold date" column and have the lookup treat anything sold in the last seven days as Sold, even if the status cell was missed.
A second before-and-after comes from tone. The first prompt version produced openings such as "We are thrilled to hear of your interest in this exceptional vehicle." Staff cut that sentence every time. Adding two sample replies written by your best salesperson to the prompt, and the rule "write like these examples", removed the problem in an afternoon.
Three smaller examples of the same pattern
Dealer with a service department. The same webhook can take a service-booking form. The AI step drafts a reply confirming the request and asking for the registration and mileage, and a router sends it to the service desk's mailbox. Sales and service never share a draft queue.
Dealer who gets photo requests. Many leads ask "can you send more photos". Add a route that matches the phrase and sends a draft linking to the vehicle's gallery page with no AI wording needed. A plain template is better than an AI draft when the answer is always the same.
Lead goes quiet. A second Zap watches the CRM for contacts with no reply 48 hours after the first draft was sent and creates a task for the salesperson, with a suggested follow-up. For ideas on the wider follow-up cycle, see the 30-day plan for a small lot.
How leads slip through: four failures to test for
- Duplicates. A buyer enquires on two marketplaces and receives two drafts. Match on email address or phone number before creating the CRM record, and stop the second run if a draft was created in the last hour.
- The sold car. The sheet was updated after lunch but the lead arrived at noon. Include a "last updated" stamp in the sheet and make the draft say "I will confirm it is still here" when the stamp is older than a day.
- The wrong name. A name typed in lower case, or a full name in capitals, comes straight through as "Hi [first name]" in the same messy form. Tell the AI step to use the first name in sentence case, and fall back to "Hi there" if the field looks like an email address.
- Consent. A reply to an enquiry is fine, but a later marketing newsletter needs its own opt-in under data-protection law such as the GDPR if you have contacts in the EU. Keep the opt-in tick as a separate field and never put someone on a marketing list because they asked about a car.
Checking it works after week one
Do four things in the first week and one every week after.
- Submit ten test leads yourself, including a sold car, a part-exchange, an empty message and one with a typo in the stock number.
- Read every draft in full for the first fifty real leads.
- Compare timestamps: lead arrival against draft created against email sent. The second gap is where staff slow down.
- Check the automation history for errors. Zapier pauses a Zap automatically if 95% of its runs error over seven days, so an unnoticed fault can silence the whole flow.
- Every Friday, sample 20 sent replies and tick: right car, right price, no invented offer, no promise nobody kept.
If you would rather map this with someone who has set it up before, that is what my AI implementation consultation covers, starting from your CRM and lead sources via the contact page. For a broader view of dealership AI, phone answering costs are the natural next comparison.
Lead-reply automation: what dealers ask next
Should the AI send the first reply without a person checking it?
Not at the start. Let the AI draft and a salesperson press send for the first month. A draft that sits in the inbox still beats a lead that waits two hours. Once the weekly sample shows no wrong prices, wrong cars or invented offers, you can auto-send only the plain acknowledgement.
What if a lead arrives at night?
Send a short acknowledgement automatically that says when a person will reply, and queue the full draft for the first shift. Never promise a time you cannot keep. A line such as 'a member of our team will reply by 9am' works only if somebody really does.
Can the AI quote a price or finance figure?
No. Prices, finance rates, monthly payments and trade-in values must come from your own stock system or a person. Pass the advertised price as a field if you wish, and tell the AI to repeat it exactly, never calculate or round it.
Which is cheaper, Zapier or Make?
For this volume Make is usually cheaper per run, but Zapier is quicker to set up if nobody on the team has built automations before. Price both against your real lead count, because the cost depends on steps per lead, not on the tool's headline price.
Further reads
- Can an AI Voice Agent Book Car Dealership Service Appointments? — Voice agents for service bookings rather than sales leads.
- AI for Used Car Dealerships: Honest Listings and Photo Clean-Up — Honest listings that make the replies easier to write.
- Responding to Dealership Reviews With AI: Samples and Rules — Reply wording for the reviews that follow a sale.
- AI BDC for Car Dealerships: Does a Small Lot Need One? — Whether a small lot needs a business development centre.
- How to Reply to Every Enquiry in Under Five Minutes With AI — The same speed idea outside the showroom.
- AI Marketing for Car Dealerships: A Four-Week Plan for One Lot — Marketing that feeds the lead flow you just built.
- Mia AI for Car Dealerships: Questions to Ask Before You Sign — A buyer's checklist for a small dealership weighing Mia: handoff, CRM link, after-hours cover, test calls and how to exit the contract.
- AI Tools for Car Sales: Test-Drive Follow-Ups, Quotes and Call Notes — A salesperson's routine for AI follow-ups after a test drive and tidy call notes, with prompts, samples and the lines AI must never write.
- AI for Used Car Sales: Price Each Car Against Local Listings — How to price each used car: comparable adverts into a sheet, mileage and condition adjustments, a floor, range and advertised price, and a guide cross-check.
- ChatGPT Prompts for Car Dealers: 10 for Follow-Ups and Trade-Ins — Ten short ChatGPT prompts for follow-ups, trade-in replies and reviews, each with sample output and a check to run before sending.
- How to Use AI in Car Sales: Trade-In Enquiries Before the Appraisal — An intake that gathers photos, mileage, service history and finance owing into a one-page pack, with guardrails so the AI never quotes a trade-in value.
- AI for Auto Dealerships: Where a Small Independent Lot Should Start — A decision rule for a lot of about 20 cars: why listings come first, lead follow-up second and phone answering last, and which finance topics stay human.
- ChatGPT for Car Sales: Follow-Ups, Listings and What Never to Promise — Follow-ups, listings and trade-in replies drafted with ChatGPT, plus the never-promise list, a rules block to paste and a listing audit.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Zapier and Make pricing and help pages, HubSpot pricing page, Anthropic API pricing (October 2026).