Life with AI Employees

I’ve been living with the beginning stages of having AI employees - agents. How so? Let’s talk about it.

Using AI to chat is one thing. You ask a question, get answer. Push back. Even have full fledged conversations and have it write code. Super valuable to me. But the AI couldn’t DO anything.

Next up was AI in my IDE. Now when I give it directions to code, it can code away creating new files, editing existing ones. Next level up in getting value. Love it.

Now we enter the world of AI agents. And here is where the experience starts to feel like having employees. My primary agent is the Warp terminal. It is my system admin on demand. We are well beyond chatting “how do I” and getting an answer that I type in myself. I now give the agent a task, and then the agent makes a plan and accomplishes the task. THAT - is starting to feel like having an employee.

I’m not speaking as in: now I can fire humans and replace them with AI. That’s a debate for another day. I don’t HAVE employees. I will never have employees. But with AI agents, it’s beginning to FEEL like I have employees.

This morning I borked my headless Linux box. I had unplugged the monitor to attach it to another computer and then back. Can’t do anything from the keyboard and monitor, the screen if frozen black Can’t remote in as I get the same black screen. So I ssh in, and then spin up my Sysadmin employee, warp.

Prompt: system is frozen when I remote in with NoMachine nx.

Employee Warp does a bunch of “stuff” but the screen is still frozen.

Prompt: it's still stuck, is it related to no longer being plugged into a monitor?

Employee Warp now has the missing context and fixes it for me. Now I can remote it.

I have chat software running in a docker container. I have my employee update it and add new models.

Prompt: Shut down the OpenWebUI container.
Pull the latest OpenWebUI and LiteLLM images
Update the LiteLLM config. Add Models to litellm_config.yaml (and then I listed the models)

And my Employee/Warp made it so.

Coding is so similar. I’m in the role of Product Owner, Solutions Architect and Scrum master. I give direction, guidance and oversight to my AI agent/employees. They do work that I then review.

I’m testifying - not reading the market or predicting the future. It sure FEELS like I have a staff. Not EXACTLY so, but a lot closer than “I use AI agents for productivity” captures.

Some of you will be aware that a new “AI Employee” has come into existing setting X/Twitter/Youtube on fire: Clawdbot. Coming up I’ll be playing with Clawdbot->Moldbot->OpenClaw (Anthropic threatened to sue on the name, and then two name changes followed). It’s blown up the past week or so and it purports to be your full time AI assistant. It’s scary and I’m sure people will get burned that aren’t careful. I’m not going to connect it to my bank accounts or give it access to send out emails or text messages in my name. But my initial “I’ll wait this one out and let others get burned and then refined” has changed. I’ve come up with passive activities I can benefit from. Research and the like. So — more on that later.

By passive I mean - I won’t be having OpenClaw taking actions on my behalf with humans or institutions. No reservations, no stock trades, no negotiating my cable bill down. But active research, schedule PLANNING, etc. Things where the product is just documents. I’m not sure I’ll be letting it code at first either. And I’ll have strict financial limits so it doesn’t run up thousands of dollars in llm bills.

Should be fun.