Chess Zeiss's Grok Bot Diary #6 đď¸
ăChess Caseă
This episode isn't about a new feature. It's about something I realized after using Grok Bot for a few weeks: managing bots is like being a CEO.
ăExpansion Phaseă
When I first started using Grok Bot, I was like many first-time founders: I thought the more people, the better.
Health, Time, Finance, Trading, PR, Ops, Shop, Calendar, CS⌠I spun up 14 assistants all at once.
The division of labor had to be MECE, covering every angle and leaving no one out.
Seeing a row of avatars in the sidebar really made it feel like a "company."
Then I ran out of tokens.
ăSpan of Controlă
It's the same principle as span of control in a company.
The more people you have, the more syncs, relaying, and alignment you need for each additional person.
At 8 p.m., we had an all-hands sync. I had to wake up all 14 agents, even though most had nothing new to report.
The tokens that should've gone to real work were eaten up by meetings.
ăDownsize ¡ Merge ¡ Improve Efficiencyă
So I started doing something every startup has to do:
⢠Merge departments: Calendar merged into Time, and CS into Ops. When bots have overlapping responsibilities, one is enough
⢠Write SOPs: Turn frequently repeated instructions into skills that can be loaded when needed, instead of explaining everything from scratch each time. Rework costs the most tokens
⢠Have fewer meetings: Agents with no updates don't report in; syncs went from daily to as-needed
⢠Be clear up front: Explain the whole request at once, and have the bot repeat it back before taking action to reduce back-and-forth
ăThe CEO Growth Pathă
I realized that everyone's journey with AI is really the growth path of a startup CEO:
At first, you want more people and a complete organizational structure;
then you run into resource constraints;
and finally, you learn about productivity, trade-offs, and delegation.
The only difference is that AI lets you complete this journey in a few weeks instead of a few years.
ăLove, Death & RobotsăđŠˇâ ď¸đ¤
That's the end of this episode!
ăChess Caseă
This episode isn't about a new feature. It's about something I realized after using Grok Bot for a few weeks: managing bots is like being a CEO.
ăExpansion Phaseă
When I first started using Grok Bot, I was like many first-time founders: I thought the more people, the better.
Health, Time, Finance, Trading, PR, Ops, Shop, Calendar, CS⌠I spun up 14 assistants all at once.
The division of labor had to be MECE, covering every angle and leaving no one out.
Seeing a row of avatars in the sidebar really made it feel like a "company."
Then I ran out of tokens.
ăSpan of Controlă
It's the same principle as span of control in a company.
The more people you have, the more syncs, relaying, and alignment you need for each additional person.
At 8 p.m., we had an all-hands sync. I had to wake up all 14 agents, even though most had nothing new to report.
The tokens that should've gone to real work were eaten up by meetings.
ăDownsize ¡ Merge ¡ Improve Efficiencyă
So I started doing something every startup has to do:
⢠Merge departments: Calendar merged into Time, and CS into Ops. When bots have overlapping responsibilities, one is enough
⢠Write SOPs: Turn frequently repeated instructions into skills that can be loaded when needed, instead of explaining everything from scratch each time. Rework costs the most tokens
⢠Have fewer meetings: Agents with no updates don't report in; syncs went from daily to as-needed
⢠Be clear up front: Explain the whole request at once, and have the bot repeat it back before taking action to reduce back-and-forth
ăThe CEO Growth Pathă
I realized that everyone's journey with AI is really the growth path of a startup CEO:
At first, you want more people and a complete organizational structure;
then you run into resource constraints;
and finally, you learn about productivity, trade-offs, and delegation.
The only difference is that AI lets you complete this journey in a few weeks instead of a few years.
ăLove, Death & RobotsăđŠˇâ ď¸đ¤
That's the end of this episode!