Vernon Campbell / AI agents in public
AI agents do the work. I keep the judgment.
I put AI agents to work inside real companies without handing over the keys. This is where I show what actually works, what breaks, and why a human still signs off.
Honest notes on running agents in a real business. No hype, no spam.
The approach
Human in the loop, for real.
Not a person rubber-stamping a machine they cannot actually check. The agents carry the busywork, the follow-ups, the moving of data. I keep final say on anything that matters, and every decision they make leaves a record you could go and check.
No hype. No "fire your whole team." Just careful, honest automation that a real business can trust with real consequences.
Field notes
What works, what breaks.
The honest version, from actually running this. Both sides, always.
Works
- Agents carrying the admin load so people do the work only people can do
- A backup before anything irreversible, every single time
- A human gate on the decisions that cannot be undone
Breaks
- Agents that report success without proof it actually happened
- Approvals relayed through other agents until no human is accountable
- "Fully autonomous" demos meeting a business with real stakes
What I've built
Real tools, real problems.
A few of them. Each one came out of a problem I actually hit.
Follow along
Building in the open.
The day to day, the wins and the misfires, plain language.
The list
Honest notes, in your inbox.
What I learn running AI agents inside a real business. What works, what breaks, and where a human still has to sign off. Plain language, no spam, leave anytime.