AI agents, and what we do about them

An AI agent is software you give a goal to. It plans the steps, uses tools and systems (email, calendars, databases, other software) to carry them out, and reports back when it is done or when it needs a decision. A chatbot answers; an agent acts.

A chatbot waits for a prompt and returns text. An agent takes initiative inside a boundary you set: it can look things up, fill in forms, move data between systems, and run multi-step work without a person typing each step. The model may be the same; the difference is the loop around it and the permissions it has.

XPLICIT is an innovation agency in Belgium. We spot the next wave in business and technology early, and help a small number of organisations act on it before it is obvious. Right now that wave is AI agents. The work is strategy, the operating model, and implementation support until it works in daily operations: see how we work.

Leadership teams of established organisations that have to live with the next wave rather than talk about it: banks, healthcare, mobility and industrial companies, and the occasional scale-up. We deliberately work with a handful at a time.

A term from Peter van Hees's book AI Agents: They Act, You Orchestrate. It describes the period in which software agents do much of the doing, and people set goals, boundaries and judgement. Companies organised for that period look different from companies organised around tools.

Not with a pilot. Start with the work (the Choose well phase): pick one flow that is repeatable, measurable and currently slow, redraw who decides what, then put an agent on the repeatable part with clear limits. Measure, widen, repeat. The technology is the easy half.

Neither in the usual sense. We are not a think tank that stops at the slide, and not a software shop that starts at the code. We sit between: in the room when a direction becomes a decision, and alongside the team until it works on an ordinary Tuesday.

Hofstade, near Brussels, Belgium (Ambroossteenweg 105, 1981 Hofstade). We work across Belgium and Europe, in English and Dutch.

With a conversation, then usually a leadership briefing or working session on what AI agents change for your organisation. If there is a wave worth riding, we agree a short first phase with a written outcome: the strategy and operating model a board can approve. Implementation follows only if the first phase earns it.

Yes. We work in English and Dutch, on site in Belgium and remotely across Europe. The books and most of the public writing are in English.