Evolving Bot Interactions for Efficient Task Management

After three days of use, my bots have streamlined processes, shifting from individual tasks to batch approvals. Key improvements include tiered task sorting, scheduled updates, and automated bot setups.

added by @rumanagarwal

Three days in, one of my bots refused another bot's word that I'd approved something: "I need to hear it from you here."

That's when I noticed how much my Grok Bot use has changed. Day one was "connect this". Today it's mostly "go".

Grok Bot plans, Claude Code does the work. My CX bot reviews a backlog, commits a handoff file to my shared repo and messages Claude Remote, which runs Claude Code on my Mac. The results come back to the CX bot. 4 rounds today. Claude Remote wanted my yes in its own chat, not passed along by another bot. So I say yes twice, and I'm fine with that.

What else changed:
• Tiers instead of item by item. My CX bot sorts work into "go", "check and go" and "needs a plan", so I approve a batch in one word.
• Schedules over chat. This was the biggest win as my chief of staff sends an attention list every hour, 9am to 9pm: what needs me across Slack, email and calendar. I tried it for one afternoon and asked for it again the next day. It skips the hour when nothing is new.
• Bots setting up bots. One bot spun up my data analyst with a rule I'd told it that morning already built in.
• One shared changelog. Grok Bot and Claude Code both write to it. When I asked if everything was logged, the CX bot audited itself and found 2 gaps.

I've stopped asking my bots for things. Mostly I approve what they've already lined up.