Monday Operations Brief
A weekly agent that reads the website and podcast analytics and the Kajabi platform’s figures, and emails leadership the biggest changes, green and red flags, and suggestions only when something is worth raising. Claude reads and writes; its GA4 reports and its email go through a no-AI n8n workflow.
For a businessLeadership gets a weekly read on site traffic and enrollments in their inbox, without opening a dashboard, and only hears about issues worth acting on.
- Status
- In production at my employer
- Result
- Not yet measured against an answer key. It sends every Monday, and leadership judges what it flags
- How it was measured
- Rebuilt as a Claude task over n8n; a test action runs it on real data and sends it to me only
- Biggest limit
- Its briefs aren’t scored yet; checking its figures against GA4 and Kajabi each week would be the first measure
01 · Context The problem
Every Monday the question is not what the numbers are, but which of them deserve a leader’s attention. Leadership wanted that judgment made for them, without logging into dashboards.
Just as important: they didn't want an automated email that manufactures urgency every Monday.
02 · Pipeline How it works
- A Claude task starts the week
A scheduled Claude task works out which comparisons it needs, asks n8n for each GA4 report and reads Kajabi through its API connection.
- n8n runs the report, read-only
A no-AI workflow checks each request (one of two allowed analytics properties, valid dates, at least one metric), runs a read-only report of up to 100 rows and returns flat rows.
- Read the week
Claude compares week over week for the biggest changes, green flags, red flags and anything worth optimizing, and suggests something only when it is clearly worth doing.
- Send to a fixed list
Claude hands n8n the finished email; n8n sends it from the shared mailbox to the three recipients written into the workflow. A test action sends it to me only.
03 · Choices Design decisions
Claude decides, n8n acts
The model writes the brief but can only run GA4 reports or send mail through one n8n workflow with three named actions. The workflow is never published, so its webhook has no public address; Claude runs it through n8n’s MCP server.
Allow-lists in code, not in the prompt
Only the two allowed GA4 properties can be queried and only the fixed list can receive the brief. A request for anything else fails in code, whatever the prompt says.
An email, not a to-do list
It offers suggestions and never says "we have to." The goal is an informed reader, not a task list.
Quiet weeks stay quiet
If nothing is obvious, the suggestions section is left out entirely rather than printing "nothing this week."
04 · Evidence How I tested it
Built first as a Zapier agent, through four revisions against the live accounts. Rebuilt as a Claude scheduled task over n8n, so its GA4 reports and its email are fixed by code rather than by a tool list.
A test action sends the brief to me alone, marked [TEST], so a change can be checked on real data before leadership sees it.
05 · Outcomes What it has done so far
- Sends the brief every Monday, running at my current employer without interruption.
06 · Honest notes Known limits
- Kajabi is read through its own API connection, not the n8n workflow, so the allow-lists above cover GA4 and the email, not Kajabi.
- Each report covers one date range, so a week-over-week comparison takes two calls.
- The n8n workflow has no “sent once” guard: a repeated send call would send the brief twice.
- Its briefs aren’t scored against an answer key; leadership judges what it flags.
Want to see this one run?
I'll screen-share the workflow, the test set and the results, including what didn't work.