What if we no longer understand our own work?

In an opinion piece for Trends, Geert Vromman argues that AI agents are not simply the next wave of automation. Classic automation starts from predictability: known input, fixed rules, a defined output. With an agent, you no longer describe the steps. You describe the result you want, and the agent decides how to get there. That looks a lot like delegating work to a colleague.

Delegating to an agent asks for far more explicit instructions than delegating to people. Colleagues rely on experience, company culture and feedback. An agent relies on what you tell it. Where instructions leave room, the agent fills it with its own interpretation. And when something goes wrong, only people who still understand the full process can trace where it happened.

Geert sees three consequences for organisations:

• Ownership stays with the business. IT takes care of architecture, security and access rights. Managers become responsible for the digital work in their team, alongside the people.
• Monitoring replaces checking every decision. Dashboards and error detection signal when agents behave unexpectedly.
• Autonomy follows impact. One agent gathers information, another makes recommendations, a third acts on its own below a set threshold. Above that threshold, a person decides, because the cost of a mistake is not the same everywhere.

The more AI automates our work, the more it matters that people still understand that work. Where does your organisation stand today?