Workflow Templates

Keep a tennis journal just by talking to your AI agent. Tell it about a match in your own words — "singles vs Sean, 6-3 6-2, backhand felt shaky" — and it logs a structured record (opponent, score, what worked, skill self-ratings, mood). Every Sunday it emails you a trends recap: what's improving, what's slipping, and what to drill. No spreadsheets, no forms, no Google Drive.

I built this to do my own marketing on X — and it's marketing this template right now. It scans X for tweets worth a reply — people frustrated with your competitors, and founders talking about the things you care about — drafts replies in your voice, and lets you approve and send each one before it goes out. It runs a single lean search so your X API costs stay low, and tracks your account's follower growth right on the canvas.

Put a teammate on trial. A /flywheel slack-court slash command opens a Block Kit form to file a (joking) case, then AI announces it, spins up a dedicated courtroom channel, and drags the defendant + witnesses in to face justice. After a 15-minute deliberation window (or an early gavel), AI reads the whole courtroom transcript and hands down the ruling.

Two workers for a support team running on HubSpot. Every new ticket is classified by AI — category, priority, and customer sentiment — with the priority written back to HubSpot and the urgent or angry ones pinged to your team in Slack. And every few hours, a sweep flags tickets that have gone quiet so nothing slips past its SLA.

Three workers for a sales team running on HubSpot. A weekly pipeline digest summarizes where every open deal stands and posts it to Slack. A daily stale-deal sweep flags deals that have gone quiet and nudges their owners. And every new contact gets scored by AI, written back to HubSpot as a lead status, with the hot ones pinged to your team in real time.

Turn your GitLab deploys into a reviewed, announced release flow. When a dev deploy succeeds, this worker writes an AI changelog of what's about to ship, posts it to Slack, and holds for a human to approve promoting it to production — approval plays your manual prod job in GitLab. When the production deploy then succeeds, it writes the release up in four voices — a Slack post, an X blurb, a Reddit write-up, and a dev.to technical post — auto-posts the Slack and X versions, and saves the Reddit/dev.to drafts for you to post. Any failed CI pipeline pages your Slack channel.

Reads unread emails from Gmail, summarizes them with AI, and delivers a daily digest. Great for inbox zero workflows.

Reads data from S3, applies conditional filtering to select relevant records, and writes results to a new S3 location.

Reads CSV files from Google Cloud Storage, applies schema mapping, and loads into BigQuery. Handles incremental updates.

AI reads each document in a folder and distills it into a clean, scannable brief — title, summary, key topics, takeaways, and action items — rendered as a formatted report you can read on every run. Clone and run it with zero setup on a set of sample documents, then point it at your own storage. No database to wire up, no output you never look at.