Seven Jira tickets in. One client update out.
Watch the pipeline read an agency's sprint, keep the internal notes back, write the update, catch its own mistake and hand it to a person to approve. This is a scripted replay on made-up data. The tool isn't built yet.
Step through it yourself
Collect
Reads the tickets and comments that changed since the last update, with read-only access to the tracker.
Clean
Moves internal comments into a store that the draft can learn from but never quote.
Classify
Sorts each ticket: done, in progress, waiting on the client, a scope question, or internal work to leave out.
Draft
Writes the update in plain language. Every line carries the ticket it came from.
Guardrails
Blocks unsourced lines and anything that repeats an internal comment, and flags promises, dates, money and awkward asks.
Review
A person who knows the project checks the flags and approves. They send it from their own email. The tool can't send.
The agency, client, tickets and run are made up, and the replay is the same every time. It shows how the tool is meant to work. The tool hasn't been built yet.
How does your agency write these today?
If you found this through a post, reply there. How long do client updates take each week, and who writes them?