A senior director at GitHub built 40 scheduled AI automations using the GitHub Copilot desktop app and credits them with recovering roughly an hour per day lost to manual context-switching. The automations run on a set schedule against live work surfaces: Google Calendar, email, Slack, and GitHub repos, connected via MCP servers. They are not chatbot queries. They are standing briefs that execute without prompting, every day, across all the places work actually happens.
The article is worth reading for the specifics, not the conclusion. The author details three automation categories that solve distinct leadership failure modes: a morning brief cluster that includes a Pre-Meeting Access Check to confirm doc permissions before a meeting starts, a Ship Decoder that summarizes every GitHub product launch in the prior 24 hours in plain language, and a Daily Wins Recap that runs each evening to log accomplishments before they disappear from memory. That last one feeds directly into performance review documentation, an annual pain point the author traces to a cognitive pattern, not a character flaw. The author is AuDHD and frames the entire system as compensating for a brain that is strong at pattern recognition and weak at thread-tracking, which reframes automation as accessibility infrastructure, not productivity theater.
The GitHub Copilot app is available on macOS, Windows, and Linux and supports parallel agent sessions across repositories using separate branches and worktrees. The automations tab, where all of this starts, prompted the author with six suggestions on first open after a single natural-language request. The full list of 40 automations is not published, but the article walks through enough categories to make the architecture legible. If you manage a team and your brain is currently the only system connecting your work, this is the piece to read.
[READ ORIGINAL →]