Our scheduled agents started running hours early
After a restart, some of our scheduled AI agents began firing hours ahead of time. What we paused, what we check now, and what is still open.
Insights
Notes from running our own companies.
What we learn putting sales follow-up, back office systems and AI to work, first in our own businesses and then in our clients'.
- 3 min read
- 4 min read
Five places a rule dies inside an AI agent's run
Writing rules for an AI agent is the easy part. These are the five points where they stop being followed, and the check that catches each one.
- 3 min read
What a vending machine test says about agents in your back office
In Vending-Bench, AI agents running a simple business for a long stretch drifted into failure. The cause matters for anyone handing agents recurring work.
- 3 min read
Write down what can go out without you
Every owner approves too much or too little. A short written list of standing approvals, each dated and limited, fixes both.
- 3 min read
A check that cannot fail is worse than no check at all
A check that cannot fail gets trusted anyway. We found two in our own systems this weekend. How to test the approvals in your business.
- 3 min read
The Replit database deletion, read as a memory failure
An AI agent deleted a production database during a declared code freeze. What a written company memory would have caught, and what it would not.
- 3 min read
Six questions to ask before you buy an AI agent
Gartner expects over 40% of agentic AI projects to be cancelled by 2027, and says many agents on sale are not agents. Six questions to ask in the demo.
- 3 min read
A model upgrade quietly turned our AI output into templates
Three times this summer our own AI steps fell back to generic output without an error anyone could see. A timeline, and what a model change exposed.
- 4 min read
Let AI draft everything. Let a person release it.
Most AI damage happens at the moment something leaves the building. Where to put approval gates, and why written rules alone are not enough.
- 3 min read
The AI drafts looked finished. None of them were real.
A drafting step in our own back office was quietly returning templates instead of real drafts. No error, no alert. How we found it and what changed.