Sunday afternoon. On the couch. Phone in hand.

I realized I forgot to update the sales page for the workshop.

Old me would've opened the laptop. Logged into the website builder. Found the page. Scrolled to the right section. Made the edits. Hit publish.

15 minutes minimum. Probably 30 by the time I fought with the page builder.

New me talked into my phone.

"Go to the sales page. Remove the early bird bonus. Reset the countdown timer to Saturday. Update the copy to reflect the new pricing."

Few minutes later... done. Sales page updated. No laptop. No login. No manual editing.

I sat there for a second thinking about what just happened.

I didn't write copy. I didn't navigate a website builder. I didn't fight with a drag-and-drop editor or wonder if the mobile version looked right.

I just... told my AI what I wanted done. And it did it.

I went back to watching Netflix.

Last week I showed you what it looks like to run your entire business from your phone in Cabo. (If you missed it, it's right here.)

But I glossed over something important...

The cool part wasn't the phone. The cool part was what was happening on the other end of it.

I wasn't asking AI a question and getting an answer back.

I was giving an AI agent an objective... and it went and did the work.

Massive difference between those two things.

And most people aren’t seeing it yet.

AI Isn’t Just a Fancy Search Engine

Most people think of AI as conversational AI. ChatGPT. Claude. You open a chat window, ask a question, get an answer, close the tab.

That's useful. I'm not knocking it.

But it's basically a really smart search engine with a chat interface. You type something in. It gives you something back. Then YOU go do something with that output.

You copy the email draft into Gmail. You paste the social caption into your scheduler. You take the outline and write the blog post yourself.

The AI thinks. You complete the task.

That's conversational AI.

Agentic AI is a completely different thing.

You give an agent an objective. Not a question... an objective.

• "Go update my sales page."

• "Draft a newsletter using my voice guide and last week's post as context. Then hand it off for social content."

• "Check my CRM. Flag any hot leads in the pipeline. Route them to Josh in Slack with a summary."

The agent doesn't just think about the task. It plans the task. Coordinates. Researches. Executes across your actual tools... CRM, email, social platforms, website.

I've hired hundreds of people over the last 20+ years. Most of the jobs I would have hired humans for in the past... I can now have AI do. For a fraction of the cost.

These agents think for themselves. Collaborate with each other. Execute without being told what to do.

And here's the part that f#cks with people's heads...

It does all of this without you touching anything.

One conversation with AI turns into a newsletter, social posts, images, and published content. You talk once. The system handles the rest.

(With approval checkpoints and human review, of course. I'm not insane enough to let AI publish without me reviewing it first. Yet.)

That's the shift.

And in the last two weeks... the entire industry figured it out at the same time.

Then the Big Dogs Showed Up

Three things.

Anthropic just launched Claude Computer Use. As in, the same week you're reading this. Claude can now control your Mac... click, scroll, navigate apps, fill in spreadsheets. You can even assign tasks from your phone via something they're calling "Dispatch," and it executes on your desktop while you're somewhere else entirely.

Think about that for a second.

You're at lunch and realize you forgot to send the Zoom link to your workshop attendees. And you need an automation built so every new registrant gets it automatically. You pull out your phone, tell Claude what you need, and it opens your CRM on your desktop back home... sends the email, builds the automation, and you never leave the restaurant.

I've been doing this with my custom setup since January. Anthropic just made it a product for anyone with a Pro subscription.

Two weeks ago, Perplexity launched something called Personal Computer. Runs on a dedicated Mac Mini 24/7... connects to your Gmail, Slack, GitHub, Notion, Salesforce. Their CEO put it perfectly: "A traditional operating system takes instructions. An AI operating system takes objectives."

(Sound familiar?)

And at NVIDIA's GTC conference last week, Jensen Huang called OpenClaw "the most important release of software, probably ever." Said every company will need a formal "OpenClaw Strategy." NVIDIA built an entire enterprise security layer on top of it called NemoClaw.

OpenClaw now has 247,000+ stars on GitHub. More than Linux.

1,000 people lined up at Tencent's headquarters in Shenzhen just to get it installed on their laptops.

Six months ago this was a niche project by an Austrian developer. Now NVIDIA is building enterprise tooling around it and people are literally standing in line for it.

And I'm about to show you exactly what I built.

What I Built While Everyone Else Was Typing Into ChatGPT

OK let me go deep.

In newsletter 14, I showed you Jenny... my AI COO running on a Mac Mini.

In newsletter 15, I showed you the 45 hours it took to build her and the pain points along the way.

That was version 1.0.

Version 2.0 is a completely different setup.

I added an orchestration layer called Paperclip. Open-source. And it does something my old system couldn't... it coordinates multiple agents like an actual company.

Org charts. Task management. Budget tracking. Governance.

Before Paperclip, Jenny was the bottleneck. She'd delegate to sub-agents, but they'd all report back to her. She'd synthesize. She'd decide. Every task flowed through one point.

Now the agents operate inside a structured system. Tasks get assigned, tracked, and completed without everything funneling through a single AI.

And the autonomous part is wild.

As soon as a task gets added to the system... it just passes it off. Assigns it to the proper agent. The agent does the work on the back end. And then I just get a report saying what happened.

I don't have any input. I don't need any input. It just... gets done.

And I added a new agent.

Jim.

Jim is my CMO.

He's the one I talk to most. I hop on Discord, talk into my microphone, and give Jim the big picture.

"I'm thinking this week's newsletter should be about agentic vs conversational AI. Here's the hook. Here's the angle."

Jim takes that and breaks it down into a plan.

Let me show you what this process looks like...

I tell Jim I want this week's newsletter to be about agentic AI. Give him the angle. Mention the couch story. Point him at the industry news I want woven in.

Jim puts together an outline and passes it to Murray. (Murray's my newsletter writer. He drafted the first version of what you're reading right now.)

Murray pulls my voice guide and the last few newsletters as context. First draft back within the hour.

Once I approve Murray's draft... it automatically goes to Peggy. She creates social posts, captions, and calls to action based on the newsletter content. She already knows what platforms we publish on, what formats work for each one, what kind of hooks get engagement.

Once those social posts get approved by me or Jim... they go to Gavin. He does all the image concepts and generates the graphics.

When the visuals are approved, the whole package gets pushed into AwesomeCRM for scheduling and publishing.

I don't coordinate any of that.

I talk to Jim once. The rest happens downstream.

The last piece I haven't cracked yet is getting the newsletter itself pushed to Beehiiv automatically... formatted, edited, ready to send.

Haven't tested that yet. But if I can get that working, the entire chain from "idea in my head" to "newsletter in your inbox" runs through one conversation.

(I think I can have this entire content system finished within a week. Maybe two.)

Meanwhile... Jim created a script on his own that automatically pulls my latest newsletter from Beehiiv every Sunday and stores it in Murray's knowledge base. So Murray always has the most recent context without me doing anything.

I didn't ask him to do that. He just... decided it was a good idea and built it.

(My buddy Jesse called me last week. His exact words: "I can't get your system out of my brain. It's the best thing I've seen." And Jesse's at the cutting edge of AI and virtual systems.)

Now let me be honest about something...

This is not all buttoned up.

I'm still migrating from my old setup. Some pieces are working beautifully. Some are held together with digital duct tape.

The autonomous backend is running... tasks flow through and get assigned to the right agent. But I've still got integration issues to iron out.

And there's a weirder thing I haven't fully processed yet...

I'm becoming less necessary to my own content operation.

Jim makes decisions I used to make. Murray drafts things I used to draft. Peggy handles distribution I used to manage. And increasingly... they're doing it without asking me.

I spent 24 years being the guy who does everything.

Now I'm the guy who approves things.

That's the whole point. That's what I've been building toward.

But it's a strange feeling when it actually starts happening.

(I'm not complaining.)

That's the reality of building on the bleeding edge.

I've got the workshop tomorrow and it's put me a little behind on finishing the migration. Should be wrapped up mid-next week.

Should be. We'll see.

But even in its half-finished state... this is already saving me hours every week. When I see a conversation turn into a plan inside Paperclip, and agents start spinning up, and handoffs happen between them, and the finished content shows up ready for approval...

It's f#cking wild.

I keep waiting for the moment where I stop being gobsmacked.

Hasn't happened yet.

You Don't Need This (But You Do Need the Foundation)

Look.

Every business owner wants the same thing.

Sustainable, profitable revenue growth.

That's the goal of a business. Not "content." Not "engagement." Not "followers." Revenue growth that doesn't require you to burn out getting there.

And the way you get there is by building a system that makes everything else simpler. You build it once. It greases the slide for the result you want.

Literally... you just talk into your microphone and you can do everything. Create content. Launch offers. Build websites. Publish across platforms. And when your system is hooked in tight... it can do most of it for you.

That's where we're going. All of us. Not in three years. Now.

Since I launched this newsletter in October, I've published 21 issues, 100+ social posts, and dozens of branded graphics. With zero content team. Zero employees. My total content spend last month was about $500 in API costs.

(Next month it’ll only be $100, because of efficiencies I’m implementing.)

My old company spends over $20,000 a month on content creators. And I'm outproducing them from my couch.

That's the math. And the math is only getting better.

Right now... most people are doing this manually. Creating content one piece at a time. Logging into five different platforms. Copy-pasting between tools. Spending 8-10 hours a week on content that maybe brings in a few leads.

You don't need Paperclip. You don't need a Mac Mini. You don't need 45 hours of tinkering with terminal commands.

But you do need the foundation that makes all of this possible.

Your voice guide. Your brand DNA. Your creative director. Your content repurposing system.

Without the voice work... every agent sounds like generic AI slop. Without the brand training... every graphic looks like it was made by someone who's never seen your business. Without the right specialist architecture... you're just talking to a chatbot with extra steps.

The specialists are the exact same specialists that power my agentic system. The framework transfers. The voice guide I built for Murray... same process you'll build on Saturday. Everything is platform agnostic... works in Claude, ChatGPT, whatever comes next.

Build the specialists first. The automation layer plugs in on top.

And the window for learning is shrinking fast. Anthropic, Perplexity, and NVIDIA are making damn sure agentic AI goes mainstream. In six months everyone's doing this.

The people who have their foundation built will be ready.

Everyone else will be scrambling.

The Workshop

On Saturday, March 28th, Jessica Davis and I are running a live workshop where you build your complete AI content system. Over the shoulder, step by step, with live demos and live troubleshooting.

And I'm teaching stuff I've never publicly taught before.

Skills... specific, repeatable processes you build once and save. Your AI follows the exact same workflow every time. No drift. And they get better as you refine them.

Ultra-specialists... instead of one general AI, you splinter it into platform-specific specialists that connect to your metrics and improve over time.

Projects and context management... so your AI stops forgetting everything between sessions. (Most people don't realize more context makes AI perform worse. I'll show you why.)

Connections... how to hook your AI into your actual tools so it can act on things instead of just writing about them.

I'm building my actual next newsletter during the workshop as a demo. Then showing you how I hand it to my social media specialist and creative director to turn it into a full week of content and graphics.

Jessica's handling short-form content and video... including AI video creation and the AI twin system she's been crushing with on Instagram.

You walk in with your Brand DNA (that's included as pre-work... takes under an hour). You walk out with skills, ultra-specialists, a creative director, a content factory, and connections to your actual tools.

One day. Everything built. Ready to use Monday morning.

Saturday, March 28th. 11:00am to 6:00pm Eastern. Live on Zoom.

👉  Get All the Details & Grab Your Spot Right Here.

No pressure.

Either way, next Friday you're getting another dispatch from the bleeding edge.

Until next time,

—Tim Erway

P.S. Next week I'll have a full report on how the workshop went. What worked. What broke. What surprised me. And probably a few stories about what happens when dozens of folks build their AI specialists at the same time. That's the point of building in public... you get to watch.