Estimates that go quiet need a follow-up, and a draft is enough

Most small service businesses have the same leak. An estimate goes out, the customer says they will think about it, and nothing happens. Nobody decided to drop it. Everyone was busy.

Following up is not hard. It just never reaches the top of the list. That makes it a good first workflow to automate: small, internal and easy to check.

The workflow, in one pass

Here is a hypothetical example, not client work. A landscaping company sends estimates from a spreadsheet or a quoting tool. Once a week, a script does the following:

1. Reads the list of open estimates, read-only.
2. Picks those with no reply after a set number of days.
3. Drafts a short follow-up for each one.
4. Puts the drafts in a review queue for a person.

Nothing is sent automatically. The person reads each draft, edits it, and sends it from their own email. The assistant never has send access.

Checks that matter

A draft is only useful if it is right. A few checks keep it honest:

• Does the draft name the correct job and quote the amount from the estimate, not a guess?
• Was the customer already contacted by phone or in person? Skip them if so.
• Is the estimate still open, or was it accepted or declined yesterday?

If the data is unclear, the draft should say so and leave the decision to a person. A blank is better than a confident wrong line.

Keep the access narrow

Non-regulated does not mean no obligations. Customer names, addresses and prices are still personal and business data. The workflow needs read access to the estimate list and nothing else: no write access to the quoting tool, no access to the whole inbox, no broad file permissions.

If you connect the assistant through MCP, the same rule applies. Check which tools the server exposes and remove any that write or delete. If you do not use MCP, there is nothing to audit there. The principle is the same either way: give it the least access that does the job.

The handoff

The owner should be able to explain the workflow in two sentences. Who reviews the drafts, and when? What happens if the list is empty or the lookup fails? Write those answers down on one page.

Run it beside the current habit for a few weeks. Compare the drafts with what people would have written. Adjust the wording and the day threshold. Then decide whether it earns a place.

That is the whole scope: one list, one draft, one reviewer. It will not transform the company. It will make sure that a quiet estimate gets one polite, accurate nudge.

---

I'm Jonathan Melton. I run FusionAL, focused on reviewing and integrating lower-risk internal AI workflows, one bounded workflow at a time.