Case study · Make.com + Claude

Drafts itself. Waits for one yes.

A two-scenario Make.com and Claude system for a real-estate agency: it researches the month, writes an on-brand newsletter, and emails the principal a single link to approve before anything reaches a client.

The problem

A real-estate agency wanted a recurring newsletter, the kind that keeps the office in front of buyers and sellers between deals. Every practical option was wrong. An email platform meant a monthly fee and a tool nobody wanted to own. Doing it by hand meant someone rewriting market commentary, reformatting listings, and fighting email HTML every month, which is the work that quietly stops happening by the third issue. Their existing monthly send slipped most months for exactly that reason.

They needed the polish of a designed newsletter, none of the monthly overhead, and a person still deciding what goes out under the agency's name.

How it works

The system is two Make.com scenarios that hand off to each other. The first makes the newsletter and asks. The second does whatever the answer is.

Scenario 1 · Generate and Approve · runs the first of each month

Scenario 2 · Decide and Distribute · waits on the click

The reason it is two scenarios and not one is timing. The approval can come minutes or days after the email lands, and a single scenario cannot sit running the whole time waiting for a click. So the decision half is its own webhook-triggered scenario that wakes up only when a button is pressed.

The part worth explaining

The plumbing is the easy half. Anyone can chain a few Make modules. The quality lives in the content skill, and specifically in refusing to let it be generic.

It works from a live research feed rather than a static prompt, so each issue reflects the actual month, not a template. It follows a fixed section skeleton, so the newsletter is recognisably the same product every time. And it scores its own draft against a corpus of the office's real past writing before the approval email is allowed to fire, so the voice is theirs and not a language model's default.

The skill lives in versioned storage as a reusable artifact, not as a prompt buried in one Make module. I improve the writing in one place, and every future issue inherits it, without touching the automation.

The email itself is deliberately old-fashioned: inline CSS and a table-based layout, because that is what actually renders in the mail clients people read.

Built to be trusted

Every send is gated behind a person on purpose. The agency's reputation is on each issue, so the automation writes, researches, and formats, but it never decides to publish. A human approves before anything reaches a client. I would rather ship the version I can stand behind than an unattended blast I have not earned.

Outcome

1 click
from finished draft to sent, principal-approved
Monthly
a reliable cadence for a send that used to slip most months
2 scenarios
generate and approve, then decide and distribute
R0 / mo
no new email-marketing subscription, runs on the tools they already have

The shape is portable. Any small business that wants researched, on-brand, human-approved outbound can run the same research, draft, preview, approve, distribute flow.