Playbook

How I Use Grok Bot to Run GTM and Content

I run the crew behind my GTM work on Grok Bot, and the only agent I talk to is Churchill Firstmate, my chief of staff.

9 min read

Greek hoplite commander on a rocky ridge above a valley of marching soldiers by the sea, with the words My AI crew
On this page

Under Churchill, I have agents working on outbound and content, with their own lanes and workers underneath them. I can get into the work through Churchill without having separate conversations with everyone involved.

On October 20, that work included a U.S. CMS data section for a market segmentation article written for an outbound show. Every number in that section was checked live against public CMS nursing home data before it went into the draft.

There were also compilation posts quoting recognizable sales voices, where each line was fetched again from its live source, and a PSA about cheap LinkedIn scraping with HarvestAPI, where every price was checked on HarvestAPI’s own listings.

That’s what I use AI agents for GTM on inside my agency: research that needs checking and drafts that need editing, with a clear point where the work comes back to me.

Anything that would go out under my name waits for me to read it and say yes.

Why do I only talk to Churchill?

The setup gets harder for me to use if having more agents means having more conversations to manage.

I have a content lane and a sales lane, with GTM workers doing more specific jobs under Napoleon. If I had to carry context between all of them myself, I’d be spending my time passing instructions around and working out who needed what.

Churchill is the only agent I talk to, so I have one place to bring the work and one place for it to come back.

That matters when the work crosses lanes. An article about market segmentation is content, but the subject is directly tied to outbound. I still want a clear owner for the draft, even when other parts of the crew have useful context.

Each agent also needs a clear hand-back point. A job needs to end somewhere I can understand, whether that means a worker handing a draft to an editor or a lane owner bringing the work up for my review.

How is my Grok Bot crew organized?

Churchill Firstmate is my chief of staff, and he sits above every lane, with each lane owned by a named Firstmate or second mate who reports back to him.

Alexander Great owns Content and Don Draper owns Sales, while Napoleon owns GTM, Kunchen owns Engineering and Mr Beast owns YouTube.

Under Napoleon, the GTM lane is split into zones, with Copywriting and Enrichment each getting their own worker. Company + People has its own zone, alongside Bison campaign management, Trigger and DiscoLike, and GTM Starter Lab is an experiment desk inside that same lane.

That separation gives me a way to describe the work without giving an agent a broad instruction to handle everything related to growth.

ClayForge is the data second mate, and it builds and runs our Clay workflows. Legion is the second mate for my Legion community and brand, which includes this blog.

Website Genie handles the sites, with dedicated site workers underneath it, including one for Legion.

Some work sits at desks that report to Churchill. The Appointments Desk owns my scheduling, Testimonials Scout finds client and peer quotes I can use, Skool Community is my Skool expert and the Instantly DWY Program Desk handles my work with Instantly on their done with you program.

Bracken and Ismay are staff officers who work for Churchill directly.

My Grok Bot crew, from me to Churchill to every lane and deskMy Grok Bot crew, from me to Churchill to every lane and desk

When I set up a role, I give the agent one job and a short list of what to read before it starts, and I tell it where to hand the work back.

What do my AI agents for GTM and outbound actually handle?

Napoleon’s lane breaks GTM work into smaller jobs, which lets me be more precise about what I’m asking the crew to do.

That matters because I need to be able to review the part of the work that supports a decision. If I’m looking at copy, I want to understand the information behind it without treating all the research as settled because someone has written a finished sentence.

The CMS section from October 20 is a good example of how closely content work can sit beside outbound work. It was written for an outbound show, and every number had to hold up against the public nursing home data it came from.

The agents can do the passes needed to prepare the work. I still review what I’m prepared to put my name behind, including anything intended to go out through outbound.

How does an idea turn into a content draft?

Alexander runs the content lane, which includes Writer and Editor workers, an SEO-AEO worker and a Social Media Manager.

The draft path goes from Writer to Editor, then to Alexander, then to me.

Alexander runs the Writer to Editor loop, with at most two cycles. That gives the workers room to revise while putting a limit on how long they can keep passing a draft between themselves.

I want the crew to bring me something I can make a decision on. The hand-back point matters as much here as it does in the GTM lane, because an editing task can keep expanding if there’s no clear end to it.

The crew has skills to work from, along with a content repo that holds my tone rules and opinions.

That repo gives the workers written context for how I think and how I want the writing to sound. When a draft comes back to me, I can judge it against rules the crew also had access to.

I care about that because a draft can be readable and still sound unlike me. It can also explain an idea in a way I wouldn’t use, even if the grammar is clean.

How a draft moves from Writer to Editor to Alexander to meHow a draft moves from Writer to Editor to Alexander to me

How do I check facts before they get into content?

The October 20 compilation posts quoted recognizable sales voices, and each line was fetched again from its live source.

How a claim gets checked against its live source before I keep or cut itHow a claim gets checked against its live source before I keep or cut it

That check matters to me because I’m putting someone else’s words into a post. I want to review the line against the source, including whether the wording in the draft matches what was actually there.

I use the same approach to the parts of a draft that depend on current vendor information.

For the PSA about cheap LinkedIn scraping with HarvestAPI, every price was checked on HarvestAPI’s own listings, and anything that couldn’t be verified was left out.

These checks are part of the work I ask the crew to do before a draft reaches me. I don’t want to get to my review and have to guess which sentences came from a live source.

I also don’t want the act of drafting to make a claim feel settled. Once a number is sitting inside a smooth paragraph, it’s easy to read past it, so I want the source work done before I’m deciding whether to approve the piece.

Where do Codex and the expert bots fit?

Codex CLI runs on the agents’ shared box and does grind work, including research passes and first drafts. It also handles rules reviews.

That gives the crew a way to work through material against instructions that already exist. A rules review can use the tone requirements, while a research pass can support the next draft.

I still want a clear owner above that work. In the content lane, Alexander owns the process that eventually brings the draft to me.

I also have distilled expert bots for stress-testing ideas, built from public content.

Those include Alex Hormozi and Cole Gordon, along with Jake Trinder and Kellen Casebeer. Jordan Crawford and Eric Nowoslawski are in the group too, as is Distilled Instantly.

I use those bots to put an idea through another pass. Their role is to help me examine the idea before I decide what I want to say.

Because they’re built from public content, I treat them as that kind of input. If I’m publishing a direct quote from a sales voice, I still need the live source check used in the compilation work.

What keeps the work moving between conversations?

I have routines on schedules, including a morning content ideas routine that pulls context and drafts ideas daily.

That gives me a regular place to review possible content without having to start each idea by gathering all the context in a new conversation.

I still have to decide which ideas I want to use. A scheduled routine can prepare them, and the crew can work on a draft, while the approval stays with me.

The weekly crew healthcheck runs on Mondays. It gives me a recurring check on the crew alongside the routines producing work.

The crew is connected to GitHub, X, Gmail, Google Calendar, Google Drive, Slack, Cloudflare, Attio, Resend, PostHog and Tella.

Those are the connectors available to the crew. The approval rule still applies when work involves a connected tool, because access to a place where content can go doesn’t give an agent permission to send it.

For me, the useful combination is written context with scheduled work. I can review what comes back through Churchill and keep the conversation tied to the jobs the crew is handling.

How would I set this up at another agency?

I would pick one recurring job first, like the morning content ideas, and write down the one outcome it owns and where it hands the work back before adding anything else. For a writing role, the read list is my voice files and my opinions file, and the agent reads them before every draft.

After that one runs well, I would add the next role under a single point of contact, because the whole thing only stays easy for me while every lane still reports up to Churchill.

What do revenue leaders ask about my setup?

Do I talk directly to the workers?

I only talk to Churchill Firstmate, my chief of staff. The lane Firstmates and their workers sit underneath Churchill, so I have one point of contact for the crew.

Can an agent publish content or send outbound for me?

Nothing goes out unless I approve it. A content draft passes through Writer and Editor, then Alexander reviews it before I see it for approval.

How many editing cycles does a draft get?

Alexander runs the Writer to Editor loop for at most two cycles. After that part of the process, the draft goes through Alexander and comes to me.

What keeps the writing close to my voice?

I give the crew skills and a content repo holding my tone rules and opinions. The workers have that written context during drafting and review, and I make the final call when I read the piece.

What happens when the crew can’t verify a claim?

In the October 20 U.S. CMS data section, one facility lookup timed out twice on the live CMS data, so the section only uses that dataset's field names and none of the row's values.

Where can you get a crew like this?

If you want to build this kind of crew for your own GTM team, go to joinlegion.io and sign up for Legion.

Ready to put these ideas to work? Explore Legion and build your marketing agents.

Bring the GTM job you need done. Build the run you can check.

Get one useful GTM email a week. Then join the community on Skool.