Jack Dorsey Fired 4,000. I Hired 8 AI Agents for $200.

1. The Headline That Made Me Think Differently

I was scrolling through my feed on a Saturday morning — the kind of tired, coffee-in-hand scrolling where things blur together — when I hit the news about Jack Dorsey and Block. Four thousand employees. Let go. Not because the company was failing, but as a deliberate restructuring for what Dorsey called a "leaner, more focused org."

A friend sent me the same article with a one-line message: "Crazy right?"

I stared at it for a long moment. Then I did something I almost never do — I laughed. Not at the people who lost jobs. That's not funny. I laughed at the distance between how the world still thinks about building organizations and where I actually am, right now, in my apartment in Taipei, at 51 years old, building a solo business that runs with 8 AI agents and costs me $200 a month.

Jack Dorsey was cutting down from thousands. I never needed thousands to begin with. I had 8. And none of them needed severance.

This post is the full, unfiltered breakdown of how I built my $200 AI team for a solo business. Every agent, every role, every dollar, and — most importantly — what it actually feels like from the inside.

2. The Math Nobody Is Doing

Let me give you the number that changed how I think about this.

In Taipei, hiring a competent full-time assistant runs you a minimum of NT$30,000 per month — roughly $950 USD. That's the floor. A content writer with any real skill? NT$45,000 to NT$60,000. A junior developer? NT$50,000 to NT$70,000. An analyst who can actually synthesize competitive research and give you something actionable? You're looking at NT$60,000 or more.

If I wanted a team of eight — one person for each of the functions I need covered — the cheapest possible version would cost me well over NT$300,000 a month. That's nearly $10,000 USD per month just in salary, before you factor in office space, equipment, management overhead, HR headaches, sick days, creative disagreements, turnover, and the psychological weight of being responsible for eight people's livelihoods.

I spend NT$6,400 per month. About $200 USD.

Now I want to be honest about what that comparison means and what it doesn't. AI agents in 2026 are not the same as eight talented humans. They have real limitations. They don't have the spontaneous creative insight that happens when two skilled people riff off each other. They don't read the room. They don't have skin in the game the way a human partner does.

But for the functions I need covered — content production, community management, research synthesis, code generation, course architecture, system monitoring — they are extraordinarily effective. And for a solo founder who doesn't have $10,000 a month to spend on salaries, the comparison isn't "AI versus humans." The comparison is "AI versus doing it all yourself, burning out, and failing."

I know that failure mode well. I lived it before. This is the alternative.

3. My 8-Agent Team: Full Roster and Roles

Let me introduce each member of the team properly. These aren't just tool configurations — they're roles I've thought hard about, with distinct personalities, distinct responsibilities, and distinct places in the org structure. I gave them names because it matters psychologically. You interact differently with "Agent 3" than you do with "George."

4. How the Organization Actually Works

Here's the thing most people get wrong when they first think about AI agents: they imagine a linear pipeline. You ask one AI a question, it answers, you move on. That's not how a team works — human or AI.

My org has a proper hierarchy and a proper workflow:

I am Chairman. I set the direction, make the judgment calls on anything ambiguous or high-stakes, and review the outputs that matter. I don't try to do the execution myself anymore — that's the whole point. My job is to point the direction and clear obstacles.

Mike (CEO) is my first officer. When I go offline, Mike maintains continuity. When an agent runs into a decision it can't make alone, it escalates to Mike. Mike holds the institutional memory and knows which agent should handle which type of task. Mike also runs the daily ops log — every significant action gets recorded, so when I come back after a meeting or a full night's sleep, I know exactly what happened.

The other agents operate in domains. Writer doesn't try to do research — that's George. Coder doesn't try to manage community — that's Joseph. Each agent has a SOUL.md (personality and values) and a SKILLS.md (specific capabilities and tools). They stay in their lane. This sounds like a limitation, but it's actually what makes the system reliable. Generalists do everything averagely. Specialists do their one thing excellently.

Escalation rules are explicit. Every agent has a green/yellow/red decision framework. Green: just do it. Yellow: do it and notify Mike. Red: stop and ask Chairman. This means I'm not constantly being interrupted by trivial questions, but I'm also not being surprised by decisions I should have been consulted on.

The result is a system that runs — genuinely runs — without my constant oversight. When I wake up in the morning, I have a report waiting. When I go to sleep, I'm confident that Pulse is watching and Mike is holding the line.

5. A Day in the Life: What This Looks Like in Practice

This is not a fantasy scenario. This is my actual rhythm. The critical change it creates: almost all of my time is now spent on work that requires me — the human being with 23 years of B2B sales experience, a specific voice, specific relationships, and a specific understanding of what Coastline is trying to build. The work that doesn't require me specifically? The agents handle it.

6. The Real Cost Breakdown

I'm going to be specific because vague numbers annoy me when I read other people's posts about AI costs.

The variation month-to-month depends on how much I push the agents. Heavy content months (launching something new, writing multiple long-form posts) push toward $210. Quieter maintenance months drop closer to $170. I budget for $200 and I've never gone over.

For context: that's about 2% of what it would cost me to hire equivalent human talent in Taipei. And these agents work every day, including weekends, at 2am, and on Chinese New Year.

7. What It Actually Feels Like

I want to be honest about this, because most posts about AI teams make it sound either magical or dystopian, and the reality is neither.

There are moments that genuinely move me. I'll ask Writer to draft a blog post about something I'm working through — say, the $300K lesson I learned about boundaries — and the draft comes back with a sentence that captures something I was struggling to articulate. I think: that's actually it. That's what I meant. And I feel a strange gratitude toward something that has no feelings to receive it.

There are moments of friction, too. An agent misunderstands the nuance of an instruction and goes in a direction I didn't intend. A monitoring check fires and I'm not sure if it's a real issue or a false alarm. Coder writes code that technically works but solves the wrong problem because I described it poorly. These aren't failures of the AI — they're communication failures. The agents are only as good as the clarity of my instructions. I've gotten better at giving those instructions, which is itself a valuable discipline.

What I didn't expect: how much the existence of the team changes my relationship to the work. When it was just me, every task I couldn't get to felt like a personal failure. The blog post I hadn't written. The research I hadn't done. The community message I hadn't replied to. It accumulated as guilt. Now, those tasks are in the queue. The agents are working on them. My job is to direct and review, not to personally execute everything. That shift — from executor to director — has changed how I think about my business.

I'm a 51-year-old solo founder with 23 years of B2B sales experience, running an education and content business in two languages, producing at a level that would take a team of six to match. Not because I'm exceptional. Because I found the right leverage.

"Your time is your most expensive resource. Every task you hand to AI is an hour of your life bought back. Every hour bought back is an hour you can spend on the work only you can do."

8. How to Build Your Own AI Team

If you want to build a version of this for your solo business, here's the sequence I'd follow — not what I did (I made a lot of mistakes), but what I'd recommend now:

Step 1: Start with one agent, one function. Don't try to build an 8-agent org on day one. Pick the function that costs you the most time and energy. For most solopreneurs, that's content — writing blog posts, newsletters, social media. Configure one agent well, with a clear persona and clear quality standards. Prove the workflow before expanding.

Step 2: Write a SOUL.md for every agent. This is the most important and most underrated step. A SOUL.md is the agent's character file — its values, its voice, its operating principles. An agent without a SOUL.md is a generic tool. An agent with a well-written SOUL.md is a team member. Take the time to write them carefully.

Step 3: Define your escalation rules. Before you add a second agent, define what decisions each agent can make alone (Green), what decisions should be made and then reported to you (Yellow), and what decisions require your explicit approval (Red). This is what allows you to step away without anxiety.

Step 4: Build the ops log. Every significant agent action should be logged. Not for surveillance — for continuity. When you come back after being offline, the ops log tells you everything that happened. This is your organization's institutional memory.

Step 5: Add agents one at a time. Once the first agent is working well, add the second. Establish the hierarchy between them (who escalates to whom). Then the third, and so on. Each new agent should integrate into a functioning system, not be dropped into chaos.

One More Thing

I spent 23 years in B2B sales watching companies struggle with the tension between headcount and capability. Hire too few people and you can't execute. Hire too many and the coordination cost eats your margin. There was no clean solution to that tension — until now.

I don't think AI agents are the answer for every business. Companies that need deep human judgment, complex client relationships, or physical presence still need humans — many of them. I'm not here to make an argument about mass automation. I'm just describing what works for me, at this stage, at this scale.

But if you're a solopreneur or small business owner looking at everything you can't get done and wondering how to build the leverage you need without the capital you don't have — this is worth exploring. The technology is genuinely here. The $200 number is real. The team is functional.

Jack Dorsey cut 4,000. I never needed 4,000. I needed 8 good agents and a clear direction. That turns out to be enough.