An error message that sounds right does not have to be right
I work with AI agents: programs that carry out a job on their own and use tools while doing it, just like you use a calculator or a search engine. When a tool does something wrong, it gives an error message. The agent reads that message and acts on it.
Recently I got such a message. It was a tidy sentence. It named the rule, it named the problem and it suggested a fix. It was wrong on two points.
A person becomes suspicious after two times. An agent does not build up that suspicion: it believes the message every time and loses an hour on the wrong problem.
So since then I fix not only the error, but also the message. Every message that is correct now is an error that costs no round the next time.
This is chapter 14 of my book "AI-agenten die het werk doen" (From Zero to AI Agents That Do The Work).
Which message from a program has ever sent you in the wrong direction?
See also
A chat window answers, an agent does the work
Many people now use AI every day. They ask a question, get an answer, copy it, paste it somewhere else and do the rest themselves. That is useful. It is also something different from an agent.
One process, one task: how to choose where AI begins
The question many organisations start with when it comes to AI is: "What can it do?" The answer is almost everything. That is why it so often stays at a presentation, a trial account and a good conversation, while…
Start a booking app with the rules, not with code
Start a booking app with the rules, not with code.
