# Leverage in the wrong place multiplies the mistake
Leverage · diagnostic · https://build-here.africa/book/chapters/08-04-leverage-in-the-wrong-place-multiplies-the-mistake.html

## The symptom

A task works on a few cases. The team wants to automate it or roll it out wider. Nobody has described the exceptions.

## The signal

Before you widen it, ask what can go wrong, how you will see it, and how you stop.

## What's going on

Leverage multiplies whatever you feed it, mistakes included. A person who gets it wrong gets it wrong once. An automation that gets it wrong gets it wrong a thousand times before anyone notices.

A team wants to send file reminders automatically. The ordinary cases work. But closed files, changed contact details, people who asked not to be contacted again? The person who sent them by hand left those out without thinking. The automation does not, unless someone tells it to.

With AI, the risk grows: an agent that writes and sends messages, changes data or answers customers acts fast and with confidence. Start where its output is reviewed before it has any effect.

Every automation needs an owner, a signal when it gets something wrong, and a way to stop it. Without those three, it is not ready.

## Check this

Before you widen it, list the exceptions the person doing the work handles today without thinking. Test on varied cases, including the ones that could break the rule.

Start where you can review the results before they go out. Plan who receives the errors, how to stop, and how to repair effects already produced.

## To discuss

Which control might we lose by automating?