We're a lot closer to AI agents building other AI agents than most people realize.
A great example of this is my friend Nick who sells managed AI agents for $5,000 per month per client.
The coolest part of his business model is that he's not the one building them. His personal agent, Dewey, does the actual work of building and setting up his clients' agents.
Dewey buys software with his own card using Agent Card. He checks his own email through Agent Mail. He texts Nick back and forth on iMessage through Agent Phone. He even signed himself up for Higgsfield and picked his own username: GreenBeluga1607 (lol).
And Nick's clients mostly talk to Dewey directly, in a group chat with Nick, Dewey, and the client. When clients need their agents updated, Dewey goes off, builds the solution, and reports back.
Nick literally watches customer problems get solved while he's on a walk.
Nick strongly believes (and I agree) that building is commoditized. The valuable skill is knowing WHICH tool to use for which problem. Knowing HOW to use the tools is no longer important.
Nick has never learned Higgsfield. He just knows it exists, knows the use case, and points Dewey at it.
A few people reached out after me and Nick's last interview asking: "Who pays $5,000 a month for an agent? Why wouldn't they just get a $20 Claude subscription?"
Because nobody pays $5K for a chatbot. They pay $5K for an agent that makes them money.
Something Nick does that I think is genius is that he builds the agent for the client's industry, then the client turns around and sells that same agent to THEIR customers as a new revenue stream.
It becomes a business to business to business (B2B2B) product. And when your client depends on your infrastructure for revenue, they never churn.
Here's Nick's 80/20 stack if you want to build your own:
Step 1: Pick a harness and a model.
The harness is the plumbing: Hermes or OpenClaw. The model is the brain: Grok 4.5 or GPT 5.6. Harness + model = your agent.
Step 2: Give it a computer.
Agents work best with a real desktop, not a headless sandbox. Nick and I both use Orgo (we've built seven agents on it for our own company).
Use this link plus code COREY for your first 3 days free and 20% off your first 3 months (you need both the link and the code).
Step 3: Give it an identity.
Your agent needs its own email, phone number, and card so it can sign up for tools itself. Grab all three at theagentbundle.com for about $20 a month.
Step 4: Wire up memory and connections.
Obsidian for the knowledge base. Honcho.dev for persistent memory across Slack, Telegram, and iMessage. Composio to connect Gmail, calendar, and the rest of your tools.
Step 5: Add observability.
latitude.so watches customer conversations and alerts you when someone is frustrated, so you fix problems before they escalate.
Step 6: Templatize it.
Build the ideal stack once, save it as an Orgo template, then one-click clone it for every new client. Nick's full pre-configured stack is free on his GitHub.
Step 7: Run the simple client process.
Discovery call. Proposal. Trello board with four columns (backlog, working on now, to do, to do together). Slack channel. The client context dumps everything, and your agent builds their agent from it.
That's the whole system.
One agent, set up once, that builds and maintains agents for clients paying $5,000 a month each.
Start building yours this week. Reply and tell me the first job you'd give your own Dewey. I read every response.
- Corey
P.S. Nick screen-shares the entire operation in our new interview, including Dewey building a client agent live on the call. Watch it here.
A few notes/tweets/cool things sourced from the AI community.
1)
Shannon is crushing it inside AI Operator Academy.
2)
Human value is shifting from execution to judgement.
3)
This looks like fun.
Thanks for reading the Corey's Notes newsletter. I'd appreciate it if you sent it to a friend that might find it interesting.
Be sure to check out the Build With AI podcast on Apple, Spotify, and YouTube. If you're a non-technical entrepreneur who wants to learn how to make money with AI, you'll love it.
Be back next week.
-Corey
