Billy Howell walks through the AI agent team that runs his local newsletter end to end, and the four-week plan for building the same thing yourself.
Posted
3 days ago
Duration
Format
Interview
educational
Views
42.2K
991 likes
57 · 43
Big Idea
The argument in one line.
A single Grok Bot agent team, kept to one business and staffed with a handful of role-specific bots, can run a real newsletter's research, writing, and sales end to end, replacing the work a human intern used to do.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You already run a content or media side business (newsletter, directory, small e-commerce) and want to hand off the repeatable parts to AI agents instead of hiring.
You've tried AI tools scattered across ChatGPT threads and Claude projects and want a single, mission-focused system instead of an ever-growing pile of forgotten chats.
You have roughly $200-300 a month to spend on tooling and are willing to run one project at a time rather than several at once.
SKIP IF…
You're looking for a free or near-free AI stack. Grok Bot's cost only pays off once a business is generating revenue.
You want a technical, code-first agent framework. This is a no-code, chat-based tool aimed at non-technical operators.
TL;DR
The full version, fast.
Billy Howell runs a real local newsletter, The Arlington Bagel, almost entirely through a team of Grok Bot AI agents, and this episode is his playbook for doing the same. The core mechanism: keep one project per Grok Bot account to avoid context bleed and token burn, start with a single chief-of-staff agent that audits the business and names the first three teammates, then follow a four-week arc: build the team, execute without adding agents, hire and fire based on real gaps, then automate with routines and five-line daily briefs. Grok Bot's edge over generic chat tools is that it forces one thread per bot and runs each agent on its own cloud machine, so nothing gets lost or context-polluted. The best entry points for a first project are a niche newsletter or a content directory, since the two feed each other and the agent pattern (research, editorial, sales, ops) transfers directly.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Greg frames the episode: Billy will show how he runs a real business, a local newsletter, entirely through Grok Bot agent teams, and what it takes to make the $200+/month tool pay for itself.
02:22 – 03:12
02 · Grok Bot Overview
Billy shows his agent roster for The Arlington Bagel, his 6,000-subscriber local newsletter, and explains that Grok Bot replaced his cousin, who used to research, edit, and send the newsletter as an intern.
03:12 – 06:09
03 · What makes GrokBot special
Billy argues Grok Bot's real edge is a deliberately limited set of agent avatars (about 76-88 shape/color combos), which forces one bot per mission instead of an unbounded pile of chat threads, cutting switching costs.
06:09 – 08:09
04 · Pick one project
Billy's first rule: run only one business per Grok Bot account. Mixing projects burns tokens, causes context bleed, and makes every agent worse, not just the new one.
08:09 – 11:19
05 · Creating your initial agent team
Start with a chief-of-staff agent, connect it to existing business docs (Notion, Slack, Gmail), and have it audit the business and name the first three agents to build.
11:19 – 13:06
06 · First week sprint with Grok Bot
Set up the chief of staff, get the team mission-aligned around making money, start with three teammates, and let a human make the final call on tooling decisions instead of letting agents stall on them.
13:06 – 15:51
07 · Automated Routines within Grok Bot
Week two is pure execution with no new agents. Week three fills real gaps (like an inbox agent). Only then do routines, Grok Bot's term for scheduled recurring tasks, get set up so the team keeps moving overnight.
15:51 – 17:35
08 · Agent Org Structure
Agents send short five-line reports (what shipped, what's blocked, what needs the human) up to the chief of staff, which routes tasks between teammates the way a manager would at a real company.
17:35 – 19:45
09 · Best Practices for building Agents
Billy recaps the whiteboard framework: stay mission-aligned, build the team, then run the damn ball. He introduces adversarial review loops, where a panel of agents critiques work across three rounds, moving output from roughly 50% to 90% done.
19:45 – 21:12
10 · Newsletter playbook
Billy opens his live Notion board for The Arlington Bagel and explains how any niche newsletter can start the same way: pick a niche, list good sources, and connect Grok Bot to a sending platform like Beehiiv.
21:12 – 22:24
11 · Plugins
Grok Bot's pre-built tool connections are called plugins. When a plugin doesn't exist for a tool (like Shopify), the agent can install that tool's own command-line interface and use it directly.
22:24 – 25:43
12 · Newsletter playbook pt 2
A research agent scans sources twice a week and fills Notion cards with story leads; a Make.com automation formats the final blurb consistently, which Billy argues is cheaper and more reliable than having Grok Bot write every blurb itself.
25:43 – 29:06
13 · Sales agent
A sales agent monitors the inbox for sponsor interest, prices ad slots against market rates, and builds a sales sheet. It found a local bagel shop sponsor this way, with the chief of staff reviewing outbound messages before they're sent.
29:06 – 33:34
14 · Iterating and Improving Agents
Billy is deliberately anti agent-creep, only adding a new agent when the team genuinely needs it, and running periodic chief-of-staff audits to decide what to add, remove, or turn into a routine or skill.
33:34 – 36:45
15 · Grok Bot business ideas
Beyond newsletters, Billy runs a separate $200/month Grok Bot account dedicated only to a Shopify sourcing experiment, using ImportYeti to research competitor products and margins before committing to a category.
36:45 – 38:36
16 · Site stack builder
Billy's second account also runs a simple site-builder stack (Cloudflare, Porkbun, Astro, GitHub) that let him spin up and launch a booking landing page from his phone at the beach in two to three hours.
38:36 – 41:42
17 · Directories and landing pages
The same stack is well suited to content directories: a research agent writes a new brand page as markdown each day, a coding agent publishes it, and the directory compounds into steady SEO traffic and referral income.
41:42 – 42:50
18 · One project for a month
Billy's closing challenge: pick one project, run it like a real team for a month using the four-week plan, and report back rather than spending another month evaluating new tools.
42:50 – 44:21
19 · Closing Thoughts
Greg recaps the four-week framework (build, execute, hire/fire, automate) and closes by pointing out that small agent-run ideas, like a local newsletter, can compound into much bigger ones over time.
Atomic Insights
Lines worth screenshotting.
Grok Bot's biggest constraint, a limited set of roughly 76-88 shape-and-color agent avatars, is also its best feature: it forces one agent per mission instead of an endless pile of forgotten chat threads.
Running two or more unrelated businesses in the same Grok Bot account causes context bleed that makes every agent in the account less effective, not just the new one.
A single moderately active newsletter business burned through about 90% of a week's Grok Bot token allowance, using one chief-of-staff agent plus a handful of role agents.
The recommended launch order is chief of staff first, then only three teammates: research, sales, and a platform specialist for whatever tool the business runs on.
Agents should report progress in five lines: what shipped, what's blocked, what needs the human. Longer reports just burn tokens on filler.
Running a three-round adversarial review, where a panel of other agents critiques a piece of work, took one operator's AI output from about 50% done to about 90% done.
The real bottleneck in an agent-run business is often a human refusing to make a decision. One agent sat stuck for three weeks because its owner wouldn't pick Notion over Google Sheets.
Grok Bot agents run in their own cloud virtual machines rather than on the user's computer, which is why they can keep working, and burning tokens, while the human sleeps.
Anything that becomes business-critical and repeatable, like writing a formatted newsletter blurb, should get moved out of the agent and into a fixed script or a cheaper tool like Make.com.
A working local newsletter's agent team found and closed a sponsor lead simply by monitoring the inbox and pricing ad slots, with no cold outreach required.
Cold outbound with AI agents works better in small daily batches, five new prospects added at a time, than in one 200-email blast, because good outreach still needs human review.
A niche content directory and a niche newsletter are meant to run together: the directory earns SEO traffic and referral fees, the newsletter turns warm leads into recurring sponsors.
The suggested four-week arc for any new AI agent business is: week one build the team, week two execute with zero new agents, week three hire and fire based on real gaps, week four automate with routines.
Takeaway
A four-week plan for turning one AI agent team into a real business.
WHAT TO LEARN
The system that makes an AI agent team actually run a business isn't more agents, it's fewer, sequenced correctly: one project, one chief of staff, three teammates, then automation.
02Grok Bot Overview
Running two or more unrelated businesses in the same AI agent account causes context bleed that makes every agent in the account worse, not just the new project.
A limited set of agent identities forces a one-agent-per-mission habit, which cuts the switching cost of juggling many chat threads across many tools.
03What makes GrokBot special
Start with a single agent that audits the existing business (docs, chat, email) and have it recommend the first three specialist teammates instead of guessing.
Give a general-purpose agent a narrow, platform-specific specialist (like a Shopify or Beehiiv expert) once the business depends heavily on that one tool.
04Pick one project
Spend the first week getting a small team mission-aligned around revenue, and let a human break ties on tooling decisions instead of letting an agent loop indefinitely.
05Creating your initial agent team
Hold a strict no-new-agents rule during the first execution week so the team's real gaps show up before more complexity gets added.
Only expand the team once specific, observed gaps appear, such as no one monitoring the inbox, rather than adding agents preemptively.
06First week sprint with Grok Bot
Automate a business only after a human has done the workflow manually and can write down the exact steps as a routine or script.
07Automated Routines within Grok Bot
Keep agent status reports to five lines: what shipped, what's blocked, what needs a human. Longer reports cost tokens without adding useful signal.
Route routine updates through a single coordinating agent so a human doesn't have to check every specialist agent individually.
08Agent Org Structure
Run adversarial review, where other agents critique a piece of work across multiple rounds, to raise output quality before it reaches the human.
09Best Practices for building Agents
Turn a one-off feedback thread into a reusable skill so future work gets reviewed the same way without repeating the same corrections.
10Newsletter playbook
A niche newsletter needs only a source list, a research agent, and a connection to a sending platform to get a first issue out.
11Plugins
When a pre-built tool connection doesn't exist, an agent can install that tool's own command-line interface and use it directly instead of waiting on native support.
12Newsletter playbook pt 2
Move any output that must be formatted exactly the same way every time (like a recurring blurb) out of the AI agent and into a fixed script or a cheaper automation tool, saving agent tokens for judgment calls.
13Sales agent
Give a sales or outreach agent a pricing sheet and inbox access so it can surface and price inbound sponsor interest without a cold-outreach campaign.
Send small daily batches of new outbound prospects rather than one large blast, since outreach quality depends on time for human review.
14Iterating and Improving Agents
Guard against agent creep deliberately. Only add an agent when the team genuinely needs the capability, and run periodic audits to prune or convert agents into routines.
15Grok Bot business ideas
Keep experimental or unrelated ventures (like a physical-product business) in a separate account so their context never mixes with a working, revenue-generating project.
Research competitor pricing and sourcing data before committing to a product category, rather than guessing at margins.
16Site stack builder
A lightweight site-building stack (a static-site framework plus a domain registrar and a CDN) is enough to ship a working landing page in a few hours without hiring a developer.
17Directories and landing pages
Pair a content directory with a niche newsletter. The directory earns SEO traffic and referral income, and the newsletter converts warm directory leads into recurring sponsors.
18One project for a month
Commit to one project for a full month before evaluating a new tool, since tool-shopping itself is often the thing preventing progress.
19Closing Thoughts
Small agent-run ideas, like a local newsletter, are a low-risk way to learn the team-building pattern before applying it to a bigger business.
Glossary
Terms worth knowing.
Grok Bot
An AI agent-team platform where each agent runs as its own cloud-hosted bot with a dedicated chat thread, built for non-technical operators who want a team of specialized AI teammates rather than one general assistant.
Chief of Staff (agent)
The first agent set up in a Grok Bot team. It audits the existing business (docs, Slack, Gmail) and recommends which specialist agents to build next, then routes ongoing work between them.
Routine
Grok Bot's term for a scheduled, recurring task an agent runs on its own, such as a daily or weekly status brief, so the business keeps moving without the human present.
Context bleed
Performance loss that happens when one AI agent or account is asked to handle multiple unrelated projects at once, filling its working memory with irrelevant files and instructions.
Plugin
Grok Bot's term for a pre-built connection to an outside tool like Gmail, Shopify, or Notion that an agent can use without custom setup. Unsupported tools can still be reached by having the agent install that tool's own command-line interface.
Beehiiv
An email newsletter platform used to write, design, and send the finished issue once an agent has assembled the content.
Astro
A web framework for building fast, simple static websites, used here for quick landing pages and content directories.
ImportYeti
A tool that shows what products a given company sources and imports, used here to research competitor products and margins before launching a Shopify store.
Adversarial review loop
A QA pattern where a piece of AI-generated work is critiqued by a separate panel of agents across multiple rounds before it's considered finished, raising output quality without direct human editing.
Resources
Things they pointed at.
20:11toolBeehiiv
20:11toolNotion
21:40toolShopify CLI
23:40toolMake.com
37:20toolAstro
37:20toolCloudflare
37:20toolPorkbun
34:10toolImportYeti
00:05productThe Rundown
Quotables
Lines you could clip.
02:30
“But cousin Marshall, I had to let him go because Grock Bot is just so much better. Sorry, cousin Marshall.”
sets up the whole episode in one line, human replaced by agents→ TikTok hook↗ Tweet quote
06:40
“You're gonna go into debt paying for tokens... you're gonna have context bleed that's gonna make all of your agents less effective.”
a clear warning with a concrete cost, works as a stand-alone caution→ IG reel cold open↗ Tweet quote
19:00
“At some point, you're the person that has to make the decision, not the agent.”
tight, contrarian, no setup needed→ newsletter pull-quote↗ Tweet quote
29:40
“It goes from like a 50% done output to like 90% done.”
a specific before/after number for a repeatable technique→ TikTok hook↗ Tweet quote
42:00
“This is your sign to pick something that you've been thinking about and pursue it aggressively versus trying a new tool and researching ideas for four weeks.”
direct CTA, works as a closing line for any repurposed clip→ IG reel cold open↗ Tweet quote
Topic Map
Where the conversation goes.
00:00 – 06:09denseWhy Grok Bot's constraints work for non-technical operators
06:09 – 15:51denseBuilding the first agent team (chief of staff first)
15:51 – 19:45denseAgent org structure and best practices
19:45 – 29:06denseThe newsletter business, walked through live
29:06 – 33:34steadyAvoiding agent creep and running QA loops
33:34 – 41:42denseOther business ideas: Shopify, site-builder stack, directories
41:42 – 44:21steadyClosing challenge and framework recap
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
Grockbot is here, and there's a reason why it's going viral. I mean, anyone who's nontechnical could spin up these agent teams and actually run a business end to end. Today's episode is for anyone who's interested in starting a business, running a business that is completely run by AI agents.
And we show you not just big ideas that you can run, small ideas that you can run using GrockBot agent teams that do $25.10, $25,000 a month.
I had my friend on, Billy Howell, come on the podcast and show you exactly how he runs his local newsletter business using Grok Bot agent. So this episode is for anyone who's interested in starting a business that actually has agents running the entire thing.
We're gonna show you how to use GrockBot. We're gonna show you how to spin up these agents. We're gonna show you the best practices.
We're gonna show you the business ideas that work here. Enjoy the episode. I'll see you at the end, and I can't wait to see what you build.
Thank God Billy Howell is here. And, Billy, by the end of this episode, what are people gonna learn?
You're gonna get an inside look at a business that I built and run with Grokbot, and you're gonna come away with how you can set up your own business using Grokbot.
Okay. And a little birdie told me that you think Grokbot is, like, the best place that anyone could run agent teams.
Right? So you're gonna go through the best practices of GrokBot, how you're using it. You're a guy who literally tests every single tool.
You test hundreds of tools to millions of people with the rundown. You're gonna show us how Grock Bot works.
You're gonna give us the best practices. You're gonna tell us how we can use Grock Bot to build businesses and run agent teams. And by the end of the episode, can you commit that the person that is gonna give their time to this episode have will have a better understanding than anyone on the Internet to run a business with Grock Bot.
Yeah. A 100%.
Because most of the content I see around Grock Bot is wrong. They totally don't understand why this tool is powerful, And I'm gonna just transmute all that information into people's brains because I'm so excited about this and it's all I think about.
Alright. Let's do it. Okay.
So Grockbot is this new agent team tool. And if you haven't seen it, it looks like this, Greg. Now what we're looking at here is the team that runs my newsletter.
My newsletter is called the Arlington Bagel. You can see it here on the hat. Pretty sweet.
And it goes out to 6,000 Arlingtonians every Thursday.
And I used to have an intern, my cousin Marshall, run this newsletter.
He would research it, he would put it together, he would edit it, He would man the email desk. But cousin Marshall, I had to let him go because Grock Bot is just so much better.
Sorry, cousin Marshall. What actually happened is he went to work at a summer camp where they don't have Internet.
So Grock Bot kinda saved me. And Greg, have you used Grock Bot yet? I have.
Okay. Cool. How many agents do you have over here?
I have 123456.
6. Okay. So the best thing I think about Grokbot is that you have a limited amount of agents.
I did the math on this, and I think it's about 76 or 88 possible unique agents.
Because when you set them up, you've only got this many shape and color combinations. And you can upload a custom avatar.
But the point is, it it's all built around constraints, and that's why it's so useful for me. Because if you're like me, you probably use lots of agent tools.
You probably have lots of business ideas. And what that turns into is an agent for every single task you think of, a ChatGPT thread you'll never come back to, a Clog Cowork project folder that sits there until you remember about it two weeks later.
But what I like about GrokBot is you are forced to be mission oriented because it's one thread, one bot.
It's DMs. It's like talking to a teammate. And I actually think that design thing is one of its best selling points.
Color and shape is all the human brain needs to differentiate something. So you can easily recognize all your different little teammates here versus, like, a grade just like a gray thread in chat GPT or Claude.
It's just I find this tool saves you so much on switching costs and organization. Um, so right off the bat, we're saving time using this.
Am I sounding crazy here? Talking about the design of the agents in Grokbot? Well, you're not.
I think, like, you know, a few weeks ago,
the Buzz was all around Jack Dorsey's Buzz product, which I think was such an interesting product. But I think what you know, I posted a video about it, a lot of people were just like, I don't I don't know how to how to get this started.
You know? It was a lot of the technical people liked it. But what it feels like is Grock Bot is almost like a nontechnical version of some of the features of Buzz.
So if you're nontechnical and you want an agent team and you want this almost like Slack like experience, Grokbot is a good place to start. Now I think the big question that a lot of people have is, and this is what I wanna get to by the end of this episode is, you know, it's not cheap.
It's, you know, $2,300 a a month. You know, if, yeah, how can we yeah.
How can well, you have it right there. How can we make GrokBot pay for itself? Five x, 10 x, 20 x, 50 x.
Where do we go?
Okay, Greg. You're not gonna like this, but you're gonna have to pick one project to work on with Grokbot.
Okay.
So the problem that I found as someone who uses lots of AI tools and builds lots of business ideas is you can't you can't you don't have enough tokens to build four businesses as one at once.
Uh, you're you're gonna go into debt paying for tokens. And even worse than that, you're gonna have context bleed that's gonna make all of your agents less effective.
So how Grok bot works is these guys actually run on the cloud, which is different than a lot of agents kinda team products that you use.
I don't know they spin up a compute that's the difference here. Right? Is they spin up virtual machines for your agents.
Virtual machine, they all live on one computer in the cloud. Mhmm. I, like, I'm pretty sure on that.
I actually hadn't thought. But why that's important is if I do my newsletter in here and then I also do, like, managing my Twitter, And then also, I'm sorting receipts in my Gmail.
That's just gonna add context to bloat. That's gonna add files to trip them up. It it's gonna slow you down, and it's going to ultimately burn tokens that go towards your limit.
Uh, I've noticed that the Croc bot usage plan is pretty generous.
Uh, I think in the first week, I used it pretty heavily, uh, but was able to get through week one with, like, 10% of my usage left.
And that's me running one newsletter. So, Greg, like, someone starting two new businesses and managing their email.
You're gonna hit that limit, and you're gonna be like, well, I just set $200 on fire. And you're gonna be frustrated.
So you I I think you need to pick. So can you walk us through, like okay. Your cousin quits.
You need us you need to create an agent team. Like, walk us through some of your agents, how people could think about best practices setting them up, and and and just give us the sauce.
Yeah, man. So when you start, they give you a bunch of preloaded agents that I can't recreate here, but I let I picked chief of staff from the list.
And then because I already had an existing business so if you have an existing business, uh, give it access to your existing business docs and have it do day one just like an audit research kind of plan. So I had it look at my Notion, my Slack, my Gmail, and I said, hey.
Take stock of this business, and then I need you to tell me the top three agents that you suggest we create first for this team to drive revenue.
Right? Our mission is to drive revenue. So what do we need?
And so my chief of staff went if we scroll back far enough, I can probably find it, um, but that's probably not worth doing.
It kind of took it so it took a deep look at everything and it said, okay. We're gonna need a research agent that researches the newsletter items.
So it's a local newsletter, and we give you three things to do every weekend and restaurant new restaurant news. Right?
So it was like, we need a research agent. We and then chief of staff was like, I'm gonna edit it, the first issue.
We're not gonna do an editorial bot yet. So we just started with research, sales, and then the newsletter goes through Beehive.
So it suggests that we have a Beehive expert agent that just anytime we have a question about, can we do this in Beehive? Like, how does this look in Beehive?
Stuff like that, it goes to that. So if your business is really dependent on a platform, say, you want to set up a Shopify store, which is something we'll talk about, definitely having a Shopify dedicated agent is a good idea just so your chief of staff doesn't have to be an expert in everything.
So is what I'm hearing the first step is to set up a chief of staff?
Set up a chief of staff, you need to figure out how to get your team mission aligned. And so the mission of your business should be to make money, and that's where you need to start.
Have it set up three teammates to start. Um, one of the pitfalls I see people make is having their agents make too many decisions for them.
Um, so for example, I've been struggling with where does our content live when we're writing the newsletter? Does it live in Notion database?
Does it live in local files? Do we have it in Google Sheets? Agent hasn't been able to make a choice for three weeks.
Uh, so finally, I I was just like, we're doing Notion. No more tinkering. Okay.
So for some reason, I think that's just a rabbit hole people get stuck in with AI is shopping around for the perfect solution. So at some point, you're the person that has to make the decision, not the agent.
And so then I suggest, once you build the team, just run the damn ball. Run with that team for a week and see what you can get done. So if you're starting fresh, use that team to buy a domain, set up your website, uh, help you set up a new Gmail, help you make your first sales list.
So think about your chief of staff as as the CEO of the business.
CEO is not gonna run the be able to run the business if it doesn't know how it works. So I'll give you an example.
Say you wanna set up an outbound sales agent for your newsletter and sale ad space sell ad space. Have your chief of staff do that exact task once, and then once it does it and you've reviewed it and it worked, be like, okay.
Now create an outbound bot that does that thing that you just did. Yeah. It makes sense.
Basically,
you have someone you know, you you call it chief of staff. I think of it more as like a COO type person, like someone running operations.
And you're basically saying, like, before you go ahead and go crazy creating a bunch of new agents for every single job to be done that you can think of, the move is basically to perfect it with the chief of staff or the COO, Get it to a point where it works, you're happy with the output.
Once the the output is pretty good, then you're like, okay, I've earned the right to create another agent that just focuses on that. And then you give, you know I I would imagine it's important to give that particular agent a goal as well.
So with the outbound agent, maybe it's like you need to send x amount of emails or you you know, we want y amount of, you know, visits and stuff like that. And then, you know, we haven't talked about it, but, you know, within GrokBot, there is the concept of essentially a cron job or a weekly you know, something that you can have it do on a daily uh, a routine, I think it it's called a routine on GrokBot where you can have it, uh, automated.
Right? Yeah. That's exactly
that that's where I was going to next. So you don't have to do it exactly like this, but this is how I did it. I built the team.
We kind of learned how to fly the plane. Uh, no tinkering was the week two of running it with Grok with Grokbot.
I didn't allow myself to create any new agents, not mess around with skills. I mean, maybe add some connections, but it was really about execution.
And then week three is when we expanded the team. So we found those gaps. So, okay, we need someone manning the Gmail inbox because chief of staff can't be doing that every day.
Uh, we need someone manning the the Shopify merch shop. Uh, and then comes automation. Um, I mean, can do automation at the top too after you build the original team if you have a good idea of how your business should work.
Uh, but basically what I did was I said to my chief of staff, bot in here, I was like, hey, based on how we ran this this week and our agent team, what are some routines we could set up to make progress while I am not working on this.
So I want you guys to move the ball forward while I'm sleeping automatically, and you need routines for that. Uh, so it put together a plan.
Most of these agents have, depending on their role, a daily or weekly or biweekly end of day brief.
So those are that's a really powerful tool with AI agents. Let me see if I can pull one up.
So it's got this instruction here in its routine. This is the social one. He runs my he creates my graphics for Instagram and runs the content calendar.
And so the instruction is to look at your current to dos, what's new, like basically what did you create today, what blockers do you have, and then write a quick five line report and send it to the chief of staff.
I think that you you just five lines is all you need. More than that, it's you're just gonna be burning tokens in context. Uh, and then my chief of staff bot will it'll be it's it'll pull up what's shipped, what's stuck, and what needs you.
Uh, and so I told it this too. So what's been done, what's blocked, and what do I need to do to unblock?
That's awesome. And the reason why it makes sense to have multiple agents and not just giving the chief of staff to do some of these tasks is because you don't wanna burn tokens and you don't wanna
cloud up the context. Is that right? Yeah.
Well, I mean, you need you need to think about this like it's a real business and like it's a real team. You wouldn't have the the CEO or COO, let's say, Greg, of a company, uh, going in and making Instagram social posts every day. And like a real person, it takes up mental bandwidth for that agent.
Mhmm. And that's the idea. And you can click in here and see it messaging other bots and, like, asking it for updates, which is pretty cool.
That is cool. I mean, that's like a that's what's so cool about all this. That's like a glimpse into, like, the future of work
where you have all these agents kinda, like, messaging each other, getting work done, coordinating work together. And then you're just sitting there, like, kinda updating the the North Star and the metrics and the goals and the context.
Yeah. Yeah. Exactly.
And so mostly I interface through this chief of staff bot. But if I need to go deeper on something, I I cut out the middleman like you would at a real company.
Uh, I would go right to the social person and be like, hey, uh, we need these graphics. Like, I I would give it feedback and be like, can you change the subtitles?
Can you change the background? Just little edits like that. And then you can tell it like, hey, when you're done, uh, update chief of staff when we're done here.
Uh, and it'll go report back. The chief of staff kinda routes tasks around. And theoretically,
you know, you could, in the future, create a so you know, bagel social strategist that's trained on how, like, Akra Geisenberg does his Instagram posts and says that, like, pretend you're, you know, Akraig Eisenberg.
Here's some context. Imagine he was reviewing these posts, what would he say? And then feed that context back to social, Bagel Social.
Is that how you do it? Yeah. Uh, I've been lately having
having it do QA loops. Um, not so I haven't done it in Rockbot yet, but in like Codex.
This is a general agent pattern. So what Greg's talking about is how do I automate feedback, uh, for work product that agents create?
And there's this hack basically where you say, okay. Once you've done this once, create a a panel of sub agents for GrockBot. It could be, you know, the chief of staff or just other agents on the team, and have them do an adversarial review of your work.
Do that in three rounds. And that has got that has made me so much progress on stuff that normally I would have to come back and painstakingly review and critique.
I get a much more policy. It goes from like a 50% done output to like 90% done.
Yeah. I mean, it also allows the human to be less involved which is helpful.
Yeah. So to be less involved But you you really have to run. You have to be in the weeds with the team for a couple days, a couple weeks before you can run it autonomously.
And you do that with the routines, like we said, uh, and you do it with having agents review work that you you teach them how to review like you.
So a useful thing could be, so say Greg, like, I gave all this feedback on these posts that it did.
I could say, okay, take what we did in this thread and turn it into a skill, uh, that another agent can use when it's reviewing.
So it reviews exactly like Billy. Right? That's exactly what you would do.
And so if you wanna see Do you wanna see how the newsletter actually works? Do you do you wanna talk about like the business case? Yeah.
I think yeah. Absolutely. Okay.
So I do I do because I think that
there's thousands of potential newsletter businesses that could be built and run with Grokbot.
Yeah. Thousands for sure. So this is this is a local newsletter, um, but I've run newsletters in tech and sports for different clients, and basically all you need to get started is a niche and where to get the best information to to put in the newsletter.
Right? So if you're really into, I I don't know, board games, and you know all the best board game subreddits, and the best board game review sites, and and message boards, then you can make a newsletter where you aggregate and editorialize about new board games every week.
Right? And so what you would do is you would get a Beehive account.
Mine runs all through Beehive. Uh, I like to use Notion for laying out my newsletter, and you can connect Grockbot to it.
It's really easy. By the way, GrockBot does connections way better than any other agents.
I think it it it's called plugins, right, on GrockBot?
It is called plugins. That's true. Okay.
So this is a look at my newsletter in Notion. And so in Grokbot so sometimes this is a pro tip.
If you don't see a plugin that you want to use or if, for example, you wanna do Shopify, Shopify can it it can like the built in GrockBot integration or plugin just is skills for using Shopify.
I think you can't actually do anything in your store, so you'd have to install the Shopify CLI. And so to do that, I would just say, hey, set up this command line tool, and it can use that on its own computer here that you can do at any time.
And that's because Shopify has a CLI. Right?
Yes. Yeah. But it's pretty intelligent.
If you wanna connect something, it'll just create, like, a card in the chat and be like, okay. Here's where you put your API key or click this to go sign in with OAuth. It's it's, like, really easy to use.
Very low friction. Okay. So basically, what you would do for creating a newsletter with Grok bot is you would make a list of your sources.
It it can be a Google Sheet, it can be a Notion, and then you have your research bot. Go out and look. So I do mine twice a week.
Uh, we send once a week on Thursdays, and I tell it to go look. And then, uh, it fills in these cards in Notion with, okay, this this is a new restaurant story.
Um, here's when we should publish it. Here's a a short summary.
And then if it's an event, like, here's one that's showing Shrek.
It'll look like this. And I have it set up so then when I move it to process, uh, there's a make.com automation that writes like the actual blurb for my newsletter how I want it, uh, which is a little advanced.
But you can have Grock Bot write the actual, like, news item for you, like, in the page in Notion. Yeah. Like, why even why even have a make.com automation?
So this is like a big agent thing in general is once you've got a workflow that's working for you, you wanna get rid of all the ambiguity. So I could have GrockBot write the blurbs every week and it does a really good job.
But it it it if the model changes or I don't know. Just depending on how the bot's feeling, it might write them differently because that's just how it is.
They're a black box. And so if you're running something business critical, you want to automate it as much as you can.
And that also frees up tokens for you to do more building with GrokBot. Right? So you wanna you wanna build, you wanna execute, and then and then automate.
Um, so when but you can have GrockBot automate stuff. So I can and then Greg, you're right.
I can do a good job having it make an SOP, and I can say, okay, every time you write an article, a news story, run this script that formats it exactly how I want.
And Grockbot can run a script on its computer every time. Uh, so you could do it that way. And it's just I don't know.
You can do more volume on make.com with like an OpenAI key for little
widget tasks like that. So it's a cost thing.
It is. Yeah. So like with the scale of a newsletter, I could I could waste GrokBot tokens, uh, writing news stories on say like 200 potential news stories.
Uh, or I could have GrockBot filter out and just keep the best ones and then have my make.com chat t pt automation write the little two sentence blurb.
I don't wanna waste it's like a real employee. If you have a high performing employee, I don't wanna waste it writing little two sentence summaries, uh, over, like, every week for for a 100 items that I may or may not use.
But again, I got here because I stood up the process and did it manually with Crockbot first. I had it writing those blurbs every week.
And I can give you an example. I keep saying blurbs. But this is what these events bay that it finds basically turn into, uh, and they're formatted the same every week.
There's like emoji, uh, event title, the time, and then a summary.
That's the part that I use OpenAI for, but Grokbot can do it. Okay. And then sales.
So Grockbot actually found me a sponsor this week. Uh, and it's because I connected it to my Gmail and had it just monitor for inbound.
And there's something I missed like last week with a local, actually, bagel shop owner reached out and he actually invited me and my wife and we went over there this morning and he gave us free bagels and we're talking about him sponsoring.
And that's because Grockpot found it and was like, hey, you should follow-up and you should pitch him on this exact ad slot. And that's because I set up this outbound or, uh, this inbox monitor and this, uh, sales agent.
And the sales agent, what you have to do when you're setting up the business is okay.
I need you to set up and price everything that we sell. Uh, and so in this case, it looked at what ad slots we had and how they were priced, and then it updated them based on the market and the growth of the newsletter.
And it made a sales sheet that we could show this local business. That's awesome.
That seems like a no brainer. So I'm just thinking about myself.
It's like I should have like, let's say I have any media, you know, any media property should have, like, an email, I guess.
Right? And we should have, like first, for sponsorships opportunities, reach out to sponsorships at whatever.
And then basically what you're saying is you have Grock Bot monitoring that inbox, pricing things out, sending messages. And then does that go to your chief of staff eventually that review reviews it, or it just goes like, how does the flow work?
Yeah.
It goes to the chief of staff. The sales bot will say, hey, of staff. We we're looking at, uh, this potential sponsor.
Does this email outbound message make sense? And right now, the chief of staff runs all of this by me. And so I'm kind of being like, okay, put this in Gmail and I'll edit it and send it.
Uh, and then the idea is it'll just improve over time as I get feedback. I For something like this, Greg, I believe with cold email it's quality over quantity for especially a local newsletter.
So it that's the thing that's worth your time to make little tweaks to. And the beauty of Rockbot is it kills that writer's block for starting that outbound message.
And so here's the trick if you're new to doing this kind of outbound outreach for your business even with AI. You don't wanna boil the ocean. So you don't wanna make a sales bot and say, you know, get me every car dealership in the state of New York and write a message to every sales agent there and set up three sub domains so that my email stays, uh, not in spam.
Right? What you should do is set up an automation that goes on Monday that goes, okay, your job is to add five prospects to our list.
And then on Monday, okay, pick the top three prospects and create an ad package custom to them and draft an outbound email and then send it to chief of staff.
And you'll make more progress doing that, I promise you, than doing 200 outbound emails at a time.
What about ideas for the agents? So one of the one of my one of my GrokBot agents is just an idea agent, and his his role is to come up with ideas for my other agents, like creative ideas, because I don't wanna be this the the person always prompting my agents to come up with ideas.
Like, how how do you think about that?
For for business ideas? Or Like,
I specific ideas for agents. So for example, like Oh.
Yeah. Like, for the social.
For example, Bagel Social. It's like, you know, you I think what makes humans amazing is, like, we're at least creative humans are gonna be, like, coming up with new ideas, always trying to push it forward.
Like, you know, culture changes. Therefore, like, you know, you move social content based on where culture is going. And I find that agents, you know, one of the reasons why AI slop is the thing is it's if if it's under the same loop, it just kinda follows the same loop, and you end up getting that creative and you get product that kinda looks the same.
So my question for you is, do you have any ideas around using GrockBot to come up with creative ideas?
My example is I actually went and asked it to create an idea idea generator, and it works with my other agents.
Yeah. So how do we keep it how do we keep the Grockbot agents out of the loop of not not improving?
Right? Introduce new ideas? I don't, Greg, really.
Like, I why I like Rockbot and how I use it is I only add an agent if we really, like, need it. I just I'm so anti idea creep and agent creep because that's how you burn tokens, that's how you spin your wheels, that's how you don't make pros progress.
You can do what I did at the beginning, that chief of staff audit. You can run it regularly and say, hey, look at how we worked this week.
Mhmm. What agents can we add? What can we remove?
Um, what should we turn into a routine or a skill? Uh, you can do that kind of automated improvement.
But, no, I don't I don't have a good solution for that.
Well, you're you're you're basically saying, like, let the agents cook. Let the and and basically don't like, it sounds like what you're saying is and you've been consistent in a sense, like, don't create a lot of agents.
Create a small amount of agents that focus on high quality, high value work.
Make sure that it it meets your bar. Have a chief of staff manage these agents, but only create an agent if it is mission critical.
Correct. Yeah. And that's exactly how you should talk to your chief of staff who manages your team.
What else haven't we covered around GrokBot that people need to know about?
Well, we talked about the computer a little bit, but it can access files on your computer. So I'm not sure how you I think I think it asked permission like one time and then it's allowed to.
You can get it on your phone and then it can run either on this computer or if the app is open and your your laptop's on, it can do stuff on your computer. I would say my main takeaway from that is, like, don't go against the grain. So GrockBot works well on its computer and it can read stuff on yours, but I find myself falling into this trap of, okay, you know, generate a report.
And then put it on my desktop, and then put it in Notion. Uh, so And I I I don't know if other people are like this, but you end up just working on the system more than the actual output.
So I'm kinda like I tell my agents like, okay, work. Create stuff on your computer, and then when it's done, send it to me in chat and put it in Notion.
Not like when it's done, like after I've reviewed it. That way you don't have to be like, hey, that thing we were working on yesterday, like, where is it?
Is it is it on your computer? Is it on my computer?
Is it in the cloud? Right? You're you're just creating problems for yourself.
So I think a lot of people are gonna watch, listen to this and really like, the whole agent team idea is gonna click in their heads.
Yes. I think that there's tons of different business ideas that you can build on top of this. You know, one of them is local newsletters.
What else could people be building? Any time of new any type of newsletter doesn't need to be local. Probably would be easier not to get the ball rolling on one that's not local.
That's a whole different can of worms.
Shopify, I'm I've got a different separate GrokBot account that I pay the $200 for because I'm so crazy about keeping the context separate, and it's just Shopify.
And so I'm training it to be, uh, to use the Shopify developer tools and to source products. I have a sourcing agent, a developer agent, and it's like an email agent right now.
And I'm still setting those up, but Crockbot does an outstanding job of researching products, sourcing products.
And so I'll just I What I would do is say, I have an idea for this product. I have I really like this company and its product.
I wanna do something that's in the in the same category, similar price point. And it knew to go look at import yeti, which like can just look at people's shopifies and tell you like exactly what they're sourcing and like what their margins are.
It did that on its own. I I was kinda blown away and it made these reports on each product idea and was like, here's my recommendation.
Always ask for recommendations by the way. Here like, my bulletproof system for testing business ideas with Crockbot is I wanna do this, this, or this.
So I've got three ideas. Maybe they're similar, maybe they're different. Go research each of them, figure out how I can price it, what the cost is gonna be, and then give me your top three recommendations on how to do the damn thing.
And it just works. You it your task should not be open ended when you're doing research, or you're just gonna end up doing research for two weeks.
That's fair. Okay. So you think I mean, you think Shopify that there's still opportunity in in finding products, finding audiences for those products, and then having agent teams manage that whole process.
That's, you know, that's interesting because, like, I'll let you know, on on this podcast, we talk a lot about, like, digital products, SaaS, vertical AI.
Not so much on the Shopify side, so it it isn't it's just interesting.
Yeah. I I'm not like an e com expert.
This is an experiment I'm working on, but I think it's nice to work on something physical because the cost of creating digital things is going to zero. There's no friction there.
Actually making something physical, a brand that you can just
like build over time, I think that think that's a smart thing to do. And over here, I see you have site builder stack. What do you mean by that?
Yeah. So I I just have an agent on my other account that has this simple stack of Cloudflare, um, the CLI tools.
Uh, Porkbun is where I buy domains, so it can log into that on its computer and buy a domain. Um, and then I use this thing called Astro for for it's a framework for creating static sites, and it's connected to my GitHub. And so basically, like, I needed I needed a link for people to book meetings with me, uh, and I didn't wanna pay for Calendly.
So I said, I got it set up on this stack, and I said, go embed my three different Google Calendars on this. And I was literally sitting on the phone on the beach, and it would send me, okay, here's a mock up of this.
And I would tell it, okay, make another branch and do a mock up in this style. And it will come back and through Cloudflare, it would have separate developer links that I could review the different styles in. And then when I said, okay, I like that one, It connected it to the domain that I owned, uh, and it went live all from the beach.
And I set that up in two or three hours, that that that stack, um, and I'll probably use it for a long time just for simple I need something to exist on the Internet.
It doesn't need to be perfect or pretty. So is,
you know, is that just, like, internal tools, building internal tools, or is there an opportunity to build some of these tools and sell it to other businesses?
Yeah. So what I would use this for is landing pages. So, yeah, it's kind of an internal tool.
If I wanna sell, say, a local business on being a sponsor, I could just say, hey, you know, make a landing page that speaks directly to this customer. Or directories, uh, so Astra specifically is really good at directories.
So if you if you're interested in the SEO game oh, just drew through that. This is the stack for testing and iterating, and I would just have a research agent that adds to the directory.
So if you wanna make a directory of pickleball equipment, I would set up a coding agent that uses the stack and a research agent, and every day the research agent investigates a new brand and writes a great page on that brand as markdown, and then it hands it to the engineer who posts it.
Then you would, you know, set up an agent that collects referral links and does outbound to get to get deals once you have the traffic. And you could rinse and repeat that if you're interested in doing kind of the SEO directory game.
Yeah. So I've done a few episodes on directories in the past.
I'll include the latest one with my friend Frey, who does a good job of kind of explaining why directories matter, how it's a good, like, 10 k, 25 k, 100 k a month business.
You know, what is a directory? It's basically just curated data on a website.
So it's you you're picking a topic and you're like, here's the best x y z in this topic. And in an AI world, we actually need directories more than ever because there's just too much information.
So when it's curated, it's helpful. Also, the whole, like, SEO game is changing right now.
Like, what's being you know, like, people are gonna chat GBT, and they're going to Google AI overviews to actually get, you know, con to consume content, and they need platforms like that need directories.
So I think using Grokbot as the basis for your directory is absolutely brilliant. And the beauty beautiful thing about a directory is it's a once you build a strong directory, it's a foundation, and you're getting thousands of visits a month, let's say.
And then you could say, you know what? I wanna build a Shopify store. Actually, you know what?
I wanna build a newsletter. Oh, you know what? I actually wanna build a SaaS.
Or you know what? I wanna build a mobile app. So it's actually a really good place for people to start.
Yeah. And the cool thing about a directory is it ties in with this, not a local newsletter, but niche newsletter.
So say we're doing that board game idea, you could have it build out like board games that are good for for parties could be your niche directory, and you could have a newsletter that talks about that too. Uh, you could have this on a sub sub domain or a different domain than your newsletter, and, uh, they tie into each other.
So you can monetize, uh, by linking people to the directory and have a referral link there. You could do those warm leads from those referral links could become sponsors in your actual newsletter.
I think the the lowest barrier to entry with GrokBot to start a business is probably newsletter and then directory.
I think that stack goes together really nice. Beautiful.
Alright. I think I think we covered what we wanted to cover. Is there anything anything else you wanna leave people with?
Yeah. I would do this.
Try this for a month. Set up Grockbot for one project and, uh, run it like a real team like we went over.
And let us like, let me know. Like, tweet at me.
Like, did you make progress doing this? I've I've I'd like this is your sign to pick something that you've been thinking about and pursue it aggressively versus trying a new tool and researching ideas for four weeks.
Uh, so do this. Build a Spotify or newsletter or a directory with GrokBot, and I think you'll see what I mean.
This I haven't been excited about a tool like this.
Maybe, like, all year? Maybe this is my favorite this year?
Yeah. So tweet at me.
What's so interesting about this tool is we've seen glimpse of this in other tools, but nothing has been as simple and has just, like, worked as well as GrokBot. So it's really fun.
Also, like, the value that you can actually create is massive. And I'm glad that during this episode, we covered it wasn't just about the big ideas.
Right? These are, like, little ideas that you can start today that anyone can do, and who knows where that goes?
Yeah. Maybe that turns into a big idea. I'm sure when Rowan was starting the rundown, he didn't, you know, think it was gonna get to millions of subscribers on his on his newsletter worth millions of dollars.
So, you know, you start small, you automate what you can, you keep it you a few agents. I'd love your I'd love your framework. Build team, week one.
Execute, no tinkering week two. Week three, hire fire. Week four, automate.
And and do it with one business, and and just keep iterating and creating value for customers.
I love it. Billy, thank you for coming on. I'll include links for where to follow Billy in the show notes in the description.
And I want to hear, and Billy wants to hear, what, you know, what you think of GrokBot. So in the comments section, please let us know. I read every single comment, and I and I and I reply to Moe.
So let would love would love to see you there. And thanks again, Billy. You're a legend.
Thanks, Greg. It's been a blast.
The Hook
The bait, then the rug-pull.
Billy Howell fired his cousin from running his local newsletter and replaced him with a team of Grok Bot AI agents. In this episode he opens up the actual system: the chat threads, the routines, and the four-week plan for turning one agent team into a real, revenue-generating business.
CTA Breakdown
How they asked for the click.
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Cloudflare quietly built the payment rail for AI agents to pay tiny fees for web resources, and that rail is about to spawn a whole generation of businesses selling clean data, agent-readiness, and packaged expertise.
Greg Isenberg's nine-part framework for turning Claude Code into a 24/7 AI employee, walked through live on a real med-spa idea, from workspace and memory files to scheduled routines and parallel agent sessions.
Cody Schneider returns to build two marketing agents live on screen — a cold-outbound machine that turns LinkedIn engagement into enriched leads, and an organic engine that turns internal conversations into a daily content pipeline.
Greg Isenberg gets a live, screen-shared tour of Jack Dorsey's new agent-native chat app from an early user — and presses him on whether it actually beats Slack.
00:38
01:15
02:05
02:47
03:29
04:05
04:40
05:15
05:51
07:07
07:11
08:28
09:06
09:44
10:22
11:00
12:12
12:48
13:55
14:28
15:01
16:08
16:43
17:13
18:23
18:56
19:28
20:28
20:57
21:48
22:43
23:23
24:03
24:43
25:27
26:33
27:07
27:41
28:15
29:28
30:13
30:57
31:42
32:27
33:11
34:21
34:53
35:25
35:57
37:03
37:40
38:17
39:22
39:53
40:24
40:55
41:26
42:16
42:38
43:35
43:53
34:10
48:10
48:28
Video of the Day43:59
26:28
38:44