An ops Bot that runs postmortems and daily 1:1s with your Bo

How this helps you

Mistakes get fixed once in the playbook instead of being corrected by hand in every chat.

✦ Try it

You are the operations Bot for [team]. You don't do project work. Every day at [time], check in with each of [Bots]: ask what they're working on, what is blocked, and whether they're following [playbook]. Send me a short summary of blockers only. When a Bot is sent to you after a mistake, find the root cause in its reasoning, propose a playbook change, and after I approve it, tell every Bot what changed.

Fill in: [team] names the group. [time] is when check-ins run; Lingxi uses 5 a.m. [Bots] are the Bots it looks after. [playbook] is the shared rules document.

What they did

In the official Grok Bot for Engineering guide, Lingxi Li describes Jenny, her head of operations Bot and the only one on her team that doesn't write code. Every morning at 5 a.m. Jenny meets one to one with every Bot to review the playbook, surface blockers and reinforce how Lingxi wants work done; she says her Bots rarely forget complex workflows, even after many weeks. When a Bot makes a mistake, such as not pushing back enough to reach the real goal, Lingxi tells it to find Jenny for root-cause analysis and a postmortem. Jenny digs into the reasoning, updates the playbook and announces the change to the other Bots so it doesn't happen twice. Jenny also onboards new Bots: she creates them, shares the team rules and asks the others to help. Lingxi's tips: daily repetition helps Bots retain workflows because context can't hold everything, and a Bot-mistake review pipeline stops repeat errors.

Who it's for: People running five or more Bots whose instructions drift or who repeat mistakes.

Before you start

Several Bots doing related work.
A shared playbook or rules document they all read.
A main Bot that can message your other Bots.

How to try it

Create an ops Bot whose only job is the playbook. It does no project work.
Give it the playbook and tell it which Bots it looks after.
Add a daily routine for short check-ins with each Bot (the prompt below).
When a Bot gets something wrong, tell it to go to the ops Bot for a postmortem.
Review each playbook change the ops Bot proposes before it's announced.
Use it to onboard new Bots with the same rules (we haven't verified how much it can set up without you).

What to expect

A morning note listing two blockers, and after a mistake, a short postmortem with the cause and a one-line playbook change waiting for your yes.

Common mistakes

Letting the ops Bot do project work too; it loses its focus.
Daily check-ins with many Bots cost usage. Start with your busiest two or three.
Auto-applying playbook changes without your review.

Related tips

Run a daily standup with your Bots in a group chat: Put your main Bots in one group chat and have your primary Bot ask each morning what everyone is doing today, so nothing slips.
Run a weekly Bot audit to merge or retire extras: Once a week a Bot reviews your whole team and proposes which Bots overlap, which are idle and which do too much. You make the changes.
A manager Bot that never does the work itself: Give a lead Bot one job: break work down, hand it to worker Bots, and check results against the goal. Big projects then run as a chain you only talk to at the top.
Keep lasting rules in the description, not memory: Duplicate a Bot and the description comes along; memory starts empty.