Hi all,
Most people missed the latest shift in AI. We've moved from agents that take orders to teams of agents that drive the work themselves.
Any AI plan built on "give it a task, check the result" is already a generation behind.
How We Got Here
Allie K. Miller walked through this in a recent webinar, and it's worth seeing how fast it all moved:
- Nov 2022: Chatbots. ChatGPT arrives and we start asking a machine questions.
- March 2023: Multimodal AI. It starts working with images, not just text.
- Sept 2024: Reasoning models. It slows down and thinks step by step.
- Feb 2025: AI agents. Reasoning meets tools, so it can do things, not just say things.
- End of 2025: Multi-agents. Several agents split up the work on a single job.
- 2026: Now. Teams of agents work autonomously and check in with us.
Until recently, I sent an agent out with instructions, it came back, I reviewed the work, and I sent it out again. I was the bottleneck.
Now I set the goals and the guardrails. The agents drive the work forward and check in to make sure it's what I need.

My Agentic Team
Here's what that looks like for me. I've built an entire agentic team: a Chief of Staff (the only one I talk to), six department heads with their own domains of expertise, and dozens of skills sitting under them. Allie's is built much the same way.
They work while I'm working, working out and sleeping.

How do I manage them? The same way I manage everyone else on my team: Asana.
In my last newsletter, I mentioned I'm building an AI health coach. Here's the board for that project. Most of the tickets are assigned to AI, and one only lands on me when it needs my input or feedback.

It's also hard to picture a team of agents collaborating, so we vibe coded a visual interface. I can see what my agents are doing at all times and how they're working together, and they can ask me open questions right there.

The numbers to watch are in the bottom right-hand corner. Over the past 24 hours, my agents logged 27 hours of work. By their own estimate, humans would've needed 103 hours to do the same thing.
That's about two and a half weeks of full-time work in a single day. It's insane!
My Take
There are other ways to configure this, and my rig won't be right for everyone. But it gets me at least 10x the output I'd have otherwise (I'm starting to track where that time goes).
I'm not sharing this to show off (ok, maybe a little). I'm sharing it because most people aren't scratching the surface. In a Gallup survey of 22,000+ U.S. workers late last year, just 12% used AI daily. And in a Stack Overflow survey of developers and tech professionals, 63% rarely or never let agents finish a task without stepping in.
Pick one recurring piece of work this week. Write down the goal, the guardrails and what good looks like, then hand it off.
And I'd love to learn from you. What does your agentic setup look like? Any killer tools or frameworks that multiply your output? Drop a comment and tell me.
Best,
Dr. Michael "House" Housman