The other day in a client meeting someone said: “We already have AI — our chatbot answers questions on the website.” True. And that is exactly the best way to explain why everyone is suddenly talking about agents: the chatbot answers. An agent gets things done.

The difference in one sentence

A chatbot responds to a question with an answer — then its job is done. An AI agent is given a goal, breaks it down into steps on its own, uses tools such as email, calendars, databases or internal systems, checks intermediate results and keeps working until the goal is reached. “What is our cancellation rate?” becomes: determine the rate, compare it with the previous year, flag anomalies and file the finished report. Agentic AI — AI that acts rather than merely answers — is the umbrella term for this new class of systems, and it is the reason 2025 was declared the “year of agents” in many places.

Why now?

Three developments are coming together. The latest generation of language models can plan in multiple steps instead of just formulating text. They are learning to operate tools reliably — calling programmes, querying data, passing on results. And connecting to company systems is becoming dramatically easier thanks to new standards. Each of these developments would be remarkable on its own; together they change which work can be delegated.

Stay sober — and start anyway

For all the enthusiasm: agents are not magic. They need clearly defined tasks, access to the right systems and boundaries within which they may act. The companies benefiting now do not start with the grand vision of the digital employee but with one concrete process — and learn from it how to work with this new kind of software. That is exactly what we support our clients with: from the first use case to automation that holds up in everyday operations.

Over the coming months we will take you along on this journey — through the topics, standards and controversies of the age of agents. Stay tuned.