ChatGPT for Planning, Slack for Execution: A Weekly Workflow That Sticks
Most teams don’t need another “weekly review app”.
They need a weekly routine that shows up in Slack, asks for the right check-in, adapts when people miss, and then feeds the next week’s work—without you reconciling three different lists.
That’s what this workflow is for: plan in ChatGPT, execute in Slack, and keep one shared behavior core underneath.
What you’ll learn
- A practical ChatGPT → Slack workflow for weekly rituals
- How to keep Slack threads as execution surfaces (not new “systems of record”)
- How Buffy keeps reminders coherent across channels over time
If you’re mapping this to the broader system, start here:
The pattern: one routine, two surfaces
The key boundary is simple:
- ChatGPT is where you plan and define the routine (windows, steps, outcomes).
- Slack is where the team executes (replies, confirmations, thread-based prompts).
- Buffy behavior core stores the truth (activities, history, reminders, memory).
If you do this split poorly, you get drifting definitions:
- Slack threads update a “Slack version” of the routine
- ChatGPT updates a “ChatGPT version”
- Neither one adapts the same way when the week gets messy
Buffy is designed to avoid that by making Slack a thin adapter over the same core.
Step 1: Define the weekly routine in ChatGPT
In ChatGPT, define the routine in a way Buffy can turn into activities. For example:
“Every Friday 15:00, run a weekly metrics review in #metrics. Ask for: traffic, activation, retention, and top decisions. If we skip twice, suggest a better time or mark as paused.”
Buffy turns that into:
- a
routineactivity - the schedule (Fridays at 15:00)
- the expected outputs (decisions + follow-ups)
Step 2: Execute the check-in in Slack (threaded, not spammy)
On Friday, Buffy posts in #metrics as a threaded prompt. A typical message looks like:
- A short reminder of the ritual
- The exact replies you need
- Clear exits (done / skip)
Example thread prompt:
“📊 Friday metrics review time. Reply in this thread with:
- Traffic (what changed)
- Activation (what’s moving)
- Retention (what needs attention)
- Decisions + follow-up tasks If you can’t do it today: reply ‘skip’ and Buffy will offer recovery options.”
What makes this work is that Slack is only rendering and collecting. Buffy is logging:
- who replied
- what was captured
- what slipped
- what follows next
Step 3: Use Buffy’s feedback loop to shape the next week
After the thread runs, Buffy can summarize the results and update the next week’s planning context.
Behind the scenes, it can do decisions like:
- Favor Slack threads for this team since replies are faster there.
- Detect two skipped weeks and shift the reminder time or compress steps.
- Suggest recovery options that preserve the intent of the routine, not just “ping again”.
This is the difference between an execution surface and a second system of record.
Where this fits in the rest of your multi-channel system
If you also use Telegram or other surfaces, the same principle applies:
- the core is shared
- reminders adapt based on behavior history
- adapters only normalize messages and deliver them to the right place
Two useful companion posts:
- One Behavior Core, Many Channels: How Buffy Actually Pulls It Off
- Teams use Buffy Agent together in Slack
Next step
Run your first weekly routine in Slack using the same adapter boundary: