A tint shop can use AI to draft three short quote texts over seven days: one within minutes of the request, one on day two, one on day six. Add stop rules for any reply, any "no thanks" and any booking, then reminders two days and two hours before the fitting.
The AI writes and personalises; it should never set the price or promise a film's performance. Those come from your price list, and the fitting slot from your calendar. The sequence below is built so a wrong guess has nowhere to land.
What a quote request already tells you
Most tint quote requests arrive thin: a web form with a name, a phone number and "how much to tint my car". The first job is to make the message sound as if you read the request. That depends on the few details you capture, so settle those before writing a single text.
The fields that matter for a tint quote are the vehicle (make, model, year), which windows (all around, rear only, front two, windscreen strip), the film type the customer asked about if any, and whether there is old film to remove. Old film is the one that surprises people: removal takes labour, and a quote that ignores it turns into an awkward conversation at the counter.
If your form does not ask for these, add three optional fields and one tick box before you automate anything. A follow-up text that says "for your hatchback, rear windows only" is more likely to be answered than one that says "regarding your enquiry". I do not have a statistic for that; it is what you would notice yourself on the receiving end.
The record the AI reads from
Whether you keep leads in a spreadsheet, a CRM or a phone system's contact notes, give each lead the same labelled fields. The AI is only as accurate as the record. A simple version looks like this, filled in for an illustrative lead:
first_name: [first name]
vehicle: 2019 compact hatchback
windows: rear three, rear screen
film_requested: carbon, mid shade
old_film_present: no
quote_amount: $260 (from price list row HB-REAR)
quote_sent: Tuesday 14:10
text_consent: yes, ticked on web form
preferred_days: weekdays after 4pm
Notice that the quote amount carries the price-list row it came from. If someone later asks why the text said $260, you can find out in ten seconds. That one habit prevents most arguments.
Consent comes before the first text
Marketing and follow-up texts sit under consent rules in most markets that have them, and the details differ. I cannot give legal advice here, and nor should an AI tool. What I can give you is the practical minimum that keeps you out of trouble nearly everywhere.
- Record consent at the source. A tick box on the quote form, worded plainly: "Text me about this quote and my fitting." Store the date it was ticked.
- Identify the shop in every text. Your trading name in the first line, so nobody wonders who is texting.
- Offer a stop word in every message after the first. "Reply STOP and I'll leave you alone." Honour it instantly and in every system that can send.
- Keep sensible hours. No texts before 9am or after 7pm in the customer's day. Set the send window in the tool rather than relying on memory.
- Keep the record. If anyone challenges a message, you want the consent, the text and the timestamp in one place.
If a lead filled in a form without any consent wording, or rang and gave a number, the safer first move is a phone call or email asking whether they are happy to be texted. Ask a qualified adviser what your market requires before you send promotional messages.
The three-touch sequence, day by day
Three touches over a week is enough. Fewer misses people who were busy; more starts to feel like chasing. Each text has one job and one easy reply.
| Touch | When | Job | What the customer is asked to do |
|---|---|---|---|
| 1 | Within 5 minutes of the quote going out, inside the send window | Confirm the quote arrived and that you read the details | Reply with a day that suits, or a question |
| 2 | Day 2, late morning | Answer the question they probably have (shade, warranty, how long it takes) | Pick one of two named slots |
| 3 | Day 6, early evening | Say the quote lapses soon and offer one more slot | Reply YES to hold it, or STOP |
If the quote request came in outside your send window, touch 1 waits for the next morning. If someone asked for a quote at 11pm, a reply at 8.55am the next day, mentioning their car, still feels fast.
Touch 1: the instant acknowledgement
This is the text where AI earns its place, because it can slot the vehicle and windows into a message in the time it takes you to put the kettle on. It should be short, mention the quote, and not repeat the price (the quote already has it).
Illustrative output for the lead above:
Hi [first name], it's [shop name]. I've sent your quote for tinting the rear windows on your hatchback. Any questions on shade or how long it takes, just reply here. Which day suits you best, weekday after 4pm?
Touch 2: the objection the quote did not answer
By day two the customer has usually looked at one other shop. The second text should answer the most common unspoken question for their situation. For a rear-windows-only job, it is often how long the car will be off the road. For a full-car job, it is shade and how dark is acceptable where they drive. Write these as a short list of two or three "objection lines" that the AI picks from based on the record, never invents.
Hi [first name], one thing people ask about rear-window tint: the fitting takes about two hours and you can wait or drop the car off. I have Thursday 4.30 or Friday 5.00 free. Which would you like? Reply STOP to opt out.
The "two hours" is a number you supply from your own timing, not something the AI works out. Put job durations in a small table the tool can read.
Touch 3: the polite last call
The third text carries a deadline, but only a true one. If your quotes are valid for seven days, say so. If you have a genuine gap in the diary, mention it. A fake "last slot" teaches customers to ignore you and can breach advertising rules on false urgency.
Hi [first name], your quote for the hatchback tint is valid until Monday. Reply YES and I'll hold Friday 5.00 for you, or STOP if you've gone elsewhere. No hard feelings either way.
Writing the drafts with AI, and what to fix
You can run the drafting inside a CRM workflow, a Zapier or Make scenario that calls an AI model, or simply paste the record into ChatGPT or Claude for the first fifty leads. The prompt matters more than the tool. Here is one that has the guard rails built in:
You write follow-up texts for a window tint shop. Use only the fields I give you.
Never state a price, discount, warranty term or film performance claim that is not in the record.
If a field is missing, leave it out rather than guessing.
Write touch {touch_number} of 3. Maximum 45 words. Friendly, plain, no emojis, no exclamation marks.
Start with the shop name on touch 1 only. Include "Reply STOP to opt out" on touches 2 and 3.
Offer exactly the two slots in {slots}. Return the text only.
RECORD:
{record}
Sample output the first time you run touch 2 on a lead with a missing field (illustrative):
Hi [first name], the fitting for your car's rear windows takes around 2 hours, and our ceramic film comes with a lifetime warranty. Thursday 4.30 or Friday 5.00 work? Reply STOP to opt out.
Two faults. The record said carbon film, not ceramic, and it contains no warranty term at all. The model filled the gap with something plausible. The fix is to strengthen the prompt with "if the record does not say, do not mention it", and to add a checker step: before sending, compare any film name in the draft to the record and block the send on a mismatch. This is exactly why you review the first batch by hand.
The same principle applies to everything customer-facing. For a lightweight approval set-up, see how to set up human review for AI work without slowing down.
Stop rules: when the sequence must end
Stop rules are the safety half of the sequence. Write them down once and put them in the tool, not in a person's head.
- Any inbound reply stops the automation. The lead moves to "needs a human". A person answers within the working day.
- STOP, UNSUBSCRIBE or "wrong number" ends everything. Mark the contact as opted out in every system, including any email list.
- A booking ends the quote sequence and starts the reminder sequence below.
- A phone call logged against the lead ends it too. If you spoke, the next move is yours.
- Three touches, then silence. Mark the lead "lapsed" and do not text again about this quote. A lapsed lead can receive a single seasonal message months later only if consent covers marketing.
- A quote edited after touch 1 pauses the sequence. A changed price needs a fresh touch 1, not a stale text with the wrong number.
That last rule is the one most shops forget. Quotes get revised ("can you do the front doors too?"), and the sequence keeps firing old facts. Tie the sequence to the quote's last-edited date.
The reminder sequence before the fitting
A booked fitting needs its own, separate, three messages. Do not reuse the quote texts; the customer is now a booked job.
| Message | When | Contents |
|---|---|---|
| Confirmation | Straight after booking | Day, time, address, how long, what to bring |
| Prep reminder | Two days before, late morning | Empty the boot and glovebox, clear the rear shelf, say if the car has a dash cam or parking sensor wiring near the glass |
| Morning-of nudge | Two hours before the slot | "See you at 4.30. Reply C to confirm or R to rearrange" |
The reply-by-letter trick in the last message matters. If someone replies R, the booking goes straight back to you as a task, and you can offer the slot to the next person on your lapsed list. A silent no-show costs you a bay for the afternoon; a reply of R at 2.30pm costs you almost nothing.
After the fitting, send one aftercare text in the evening. The wording about not rolling windows down, drying time and cleaning must come from your film manufacturer's own aftercare guidance, because the safe number of days varies between products. Copy their instructions into your template rather than letting an AI recall them. A few days later, the same thread can ask for a review; getting more reviews after every job with AI follow-ups covers that ask.
Where the sequence can run, and what it costs
You do not need a bespoke system. Three realistic set-ups, from simplest to most automated (list prices in USD as of October 2026; check each vendor page before buying):
- Business phone app with texting. Quo (formerly OpenPhone) lists Starter at $19 a user a month billed annually, $28 monthly. You send and schedule from a shared number; the AI drafting is done by you pasting from ChatGPT or Claude. Cost: the phone plan plus a chat assistant at around $20 a month.
- A free CRM plus a phone line. HubSpot has a free CRM tier. You log the lead, a workflow task reminds you which touch is due, and you send from the phone. It is manual but dependable for 20 to 40 leads a month.
- Automation between your form, your sheet and your texting tool. A Zapier Professional plan starts at $29.99 a month billed monthly ($19.99 billed annually) for 750 tasks, or Make from about $9 a month. The trigger is free, but every action step counts as a task, and an AI by Zapier step can use more than one task per run, so budget several tasks per text. Add a "wait" and a "check for reply" step.
The honest advice for a shop under about 40 quote requests a month is the first option, run by hand for a month, before spending on automation. You learn which texts get replies before you hard-code them.
A worked month, with illustrative numbers
Take a single-bay shop that gets 40 quote requests a month. Average job is $380. Today, say, 12 of the 40 book (30%), mostly the ones who rang. The other 28 get one email and nothing more.
Suppose the sequence reaches 24 of those 28 with consent in place. If just three of those 24 book, that is 3 x $380 = $1,140 of extra work from the month. Tooling in this set-up: Quo Starter at $28 monthly plus a $20 chat plan, about $48. Even one extra job covers the cost several times over. If none book, you have spent $48 and an hour a week, and you have learned that your price is the issue, not your follow-up. Neither figure is a forecast; the point is that the break-even is one job.
The time cost is the bit people underestimate. Reviewing drafts takes about two minutes per lead for the first 50 leads, so around 100 minutes in the first fortnight. After that, it drops to spot checks.
Four realistic ways this goes wrong
The 10.40pm text
A shop sets up its tool in the evening, tests it, and leaves the schedule on the default "send immediately". Real leads then receive texts at night. Fix: set the send window in the tool and test with your own number at 11pm to see that it holds the message until morning.
The text after "no thanks"
A customer replies "found someone cheaper, thanks" to touch 1. The automation does not read replies, so touch 2 still goes out. Any inbound message must stop the sequence, whatever it says. If your tool cannot do that, do not run an automated sequence on it.
The invented film claim
Take the example above: an AI added a warranty. Claims about heat rejection, UV blocking or warranty length belong in a locked list of approved sentences. The AI may choose among them; it may not write new ones.
The wrong vehicle
The form allows free text, and one lead types "my truck and my wife's SUV". The AI picks one and quotes the other. Where the record is ambiguous, the draft should become a task for a person, not a send. Add an "unclear fields" check to your workflow that routes any lead with a vehicle field longer than about 40 characters to manual review.
Checking it worked after four weeks
Look at five numbers, once a week, in a spreadsheet with a row per lead:
- Reply rate by touch. If touch 1 gets few replies, your first line is too generic. If touch 3 gets none, drop it and test a shorter gap between touches.
- Quote-to-booking rate for leads who got the sequence against those who did not (a fair comparison needs a few weeks of both).
- Opt-outs. A handful is normal. A spike means the wording or the timing is wrong.
- No-shows and late cancellations before and after the reminder sequence.
- Draft edits. How often a person changed the AI's text. When this drops near zero over fifty leads, you can loosen review.
To compare, keep a control: for two weeks, hold back every fifth lead from the sequence. Small numbers are noisy, so treat anything under about 30 leads per group as a hint, not a result. If you would rather map the sequence against your own quoting steps with someone, that is what my AI implementation consultation is for.
Once texts are working, the same record can drive the rest of the job. A shop that already logs quotes carefully can turn voice notes from the bay into tidy job sheets, as in turning on-site voice notes into job sheets and quotes with AI, and decide whether an answering service is worth it using the tint shop receptionist cost guide.
A one-page template to copy
Pin this above the desk. Each line is something the AI cannot decide for you.
- Send window: 9am to 7pm, customer's time. Quote validity: 7 days.
- Touch 1 within 5 minutes; touch 2 on day 2 at 11am; touch 3 on day 6 at 5.30pm.
- Approved sentences list: shade options, durations, warranty wording, aftercare lines.
- Stop on: any reply, STOP, booking, logged call, quote edit.
- Reminders: confirmation at booking, prep at two days, nudge at two hours with C or R.
- Review: a person approves the first 50 sends; then spot-check one in ten.
Tint shop texting questions
Should the AI send the texts without anyone reading them?
Not at first. Let the AI draft each message from the quote record, and have a person approve the first fifty or so sends. Once the drafts are consistently correct on vehicle, film and price, you can let the first reply go automatically and keep every later touch under review.
What if a customer replies in the middle of the sequence?
Any reply ends the automated sequence for that lead. A person answers, because replies often contain a question the AI should not guess at, such as film warranty or whether a windscreen strip is allowed. Re-enter the lead only if the conversation goes quiet again.
Is a reminder text really worth sending for a fitting?
Yes. A fitting blocks a bay for hours, so a no-show costs real money. A text two days before and a short one on the morning does two jobs: it confirms the slot and tells the customer how to prepare the car, which also saves time on the day.
Can I text a lead who only filled in a web form?
Only if the form made clear that they agreed to receive texts about their quote, and you keep a record of that consent. If the form was silent, reply by email or phone for the first contact and ask for text permission there. Ask your adviser what applies to your business.
Further reads
- How Much Does It Cost to Automate Lead Follow-Up With AI? — What automating lead follow-up costs once you outgrow manual texts.
- Quoting Custom Jobs Faster With AI in a Small Workshop — Quote the awkward jobs that do not fit a price list.
- How Painters and Decorators Can Quote From Photos With AI — Quote from customer photos before anyone visits.
- Following Up Unapproved Repair Estimates With AI — Chase unapproved estimates with the same stop rules.
- AI for Moving Companies: Chasing Quotes and Confirming Move Dates — A nine-day quote chase, three confirmation messages and a moving-day checklist, all drafted by AI from one row of facts per job.
- Tyre Sales Promotion Ideas: Seasonal Offers Written With AI — Six seasonal tyre promotions with the margin sums done first, then AI prompts and sample copy for texts, email and posts, plus the claims to cut.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Quo (formerly OpenPhone) pricing page, Zapier pricing page, HubSpot pricing page, and your own film manufacturer's aftercare guidance.