# The manager who tried it first-hand

> I advise companies to have their leaders try AI first. So I started with myself.

Author: Aleš Nypl · https://engineloop.cz/ales-nypl

At the start of the year I was asking what “running in a terminal” actually meant. Today I have agents working for me that handle all kinds of work in different roles. They use skills, subagents and a knowledge base, and they have documented rules, procedures and their own organisation of work and time. I manage them by those rules. I wasn't a programmer then and I'm not one now. I was a manager with twenty years' experience and a good deal of curiosity.

## What I learned first-hand

- **You can't delegate accountability to a tool.** Even when AI does the work, I answer for it. That changes how I brief the work.
- **The hardest part is the rules and the system. The technology is surprisingly accessible.** Who is allowed to do what, who approves, and what happens when something goes wrong. That's management work, and no engineer can do it for you.
- **You learn fastest on your own problem.** My own email, calendar, documents or apps taught me more than any training course.

## My advice to companies

- Have the leadership team build one small thing for themselves before they start telling teams to use AI.
- For every task, decide the point where a person makes the call.
- Write mistakes down and turn them into rules.

You don't have to be able to code. You do have to want to manage.