Estimate Follow-Up Agent

Every unapproved estimate gets chased on a schedule, not when someone remembers.

Beginner~3 hours to build$0–30/month~3 hours/week saved

Every estimate that doesn't get approved on the spot goes into a queue this agent watches. Day 2, it drafts a friendly text reminding the customer their recommended work is still on file. Day 5, with no response, it drafts a more direct follow-up restating the actual consequence of waiting — a fluid getting worse, an inspection item — without fear-mongering. Day 10, it drafts a close-out message and marks the estimate cold if there's no reply. Every draft lands in a queue for your advisor to skim and send — nothing goes out untouched. A stack of forgotten estimates becomes a working pipeline that recovers work you already earned the right to do.

What this agent takes off your plate

  • Every unapproved estimate gets a scheduled follow-up instead of falling off the desk
  • Follow-ups escalate gently over 10 days instead of one text that never gets repeated
  • Safety-critical declined items are flagged for a phone call, not left to a text thread
  • A visible log of which estimates converted, stalled, or went cold
  • Service advisors spend minutes approving drafts instead of remembering who to call back

The stack

Claude or ChatGPTZapier (free tier)a shop estimate export or Google SheetSMS via your existing texting tool

Real costs & ROI

  • ChatGPT or Claude (free tier)$0/mo
    plenty for a single advisor drafting a handful of follow-ups a day
  • Zapier (free tier)$0/mo
    100 tasks/month covers roughly 3 follow-ups/day, typical for a 4–8 bay shop
  • Sending texts through your phone or existing texting tool$0/mo
    most shops already text customers manually — no new tool required to start
  • OpenPhone or Zapier's SMS integration (optional)$20–30/mo
    worth adding once you want the sending automated too, not just the drafting

The math

Shops typically lose 20–30% of written estimates to nothing more than "we got busy and never circled back" — not a real no. If your average approved estimate is $400 and you write 40 a month, even a 10% recovery from consistent follow-up is $1,600/month in work you already diagnosed and quoted. Doing that follow-up manually — checking who's pending, drafting a text, deciding on tone — runs a service advisor 2–3 hours a week, or $260–360/month at $30/hour. This agent turns that into 10–15 minutes a day of approving drafts. Even at the top of the cost range, you're spending under $30/month against thousands in already-quoted work sitting in a drawer. Breakeven is the first recovered estimate.

Build steps

1

Write your follow-up voice and tier plan

Before automating anything, decide what each follow-up should say and when it should escalate. This alone is worth doing even if you stop here.

Data you need

Pull 5–10 recent unapproved estimates — a mix of small maintenance items and bigger repairs — so the AI has real examples to work from.

Prompt
I run an independent auto repair shop called [SHOP NAME] in [CITY]. I want to build a 3-touch follow-up plan for estimates customers haven't approved yet.

Here are some recent unapproved estimates, with dollar amount and what the work was: [PASTE 5–10 EXAMPLES]

My shop's tone: [e.g., straightforward and honest / warm and low-pressure / efficient and to-the-point]
Things I never want to sound like: [e.g., pushy, alarmist, "used car salesman"]

Please build me:
1. **A day-2 touch** — friendly reminder, assumes the customer is just busy, no urgency language, ends with an easy yes/no
2. **A day-5 touch** — slightly more direct, restates the specific consequence of waiting for THIS type of work (safety item vs. wear item vs. convenience item — treat these differently), still no pressure tactics
3. **A day-10 touch** — a clear close-out message: "want us to go ahead, or should we release the estimate?" Not guilt, just clarity
4. **A tone rule for safety-critical items** (brakes, steering, tires) vs. wear items (belts, fluids) vs. convenience items (AC, cosmetic) — safety items should sound more direct sooner, without ever sounding like a scare tactic
5. **A length rule** — SMS under [CHARACTER LIMIT] characters, plain language, no jargon like "OE" or part numbers
6. **A "stop the sequence" rule** — if the customer replies at any point, the sequence pauses and a human takes over

Save this as my follow-up voice guide — every future prompt will reference it.
2

Set up your estimate tracking sheet

Give the automation something to watch — a simple list of pending estimates with dates. Be honest with yourself about the source: Tekmetric, Shop-Ware, and Mitchell 1 do not offer a reliable "new unapproved estimate" trigger in Zapier or Make, so for most shops the practical path is a daily CSV export of open estimates that you paste (or upload) into a Google Sheet. Add two columns the automation will need: a "consent to text" column (yes/no, with the date the customer agreed) and a "last touch" column so the same estimate is not chased twice in a day.

Data you need

Either export unapproved estimates from your shop management system daily, or keep a simple Google Sheet: customer name, phone, vehicle, estimate amount, work description, date quoted, safety flag (yes/no), consent to text (yes/no + date), last touch date, status.

Agent Pack

Unlock the Auto Repair Shop Agent Pack.

One payment unlocks all 3 blueprints in this pack — build steps, the complete system prompt, guardrails, a test checklist, and the maintenance routine — for $29.

See the whole pack

one-time, unlocks all 3 blueprints in this pack on this browser. Secure checkout via Stripe.

More in this pack

More for Auto Repair Shop

Service Advisor Agent

Tech notes become customer updates a non-mechanic actually understands.

Intermediate~4 hours$20–40/month~4–5 hours/week saved

Review & Reputation Agent

A review request after every completed job, and a reply drafted for every review that comes back.

Beginner~2 hours$0–20/month~2 hours/week saved