Maxxton builds two AI agents for holiday parks and opens an Amsterdam hub to do it

Maxxton has opened an AI hub in Amsterdam and is building two agents for outdoor hospitality: Max, which works for staff, and Ton, which works for guests. The claim that matters is that Max does not just answer questions about the system. It acts inside it.

That distinction is the whole argument. Most PMS vendors shipping AI this year have shipped a chat window that explains where a button is. An agent that can create a reservation is a different kind of product, with a different kind of risk.

What Max does

According to Maxxton, Max can create reservations, apply changes to them, summarise the day's priorities for whoever is logged in, give guidance on the system, and recommend relevant add-ons.

It is permission-aware, which means a front office employee and a park manager asking the same question get different answers scoped to what each is allowed to see and do. For a business where seasonal staff turn over constantly and access control is the thing that stops a summer hire reading the owner ledger, that is the right design choice.

The numbers Maxxton gives are about onboarding. It puts training time for a new employee at roughly six weeks before Max and about an hour after. At De Krim, a park using it, field workers' questions to the back office reportedly fell from around eight a day to one.

Ton, the guest-facing agent, is described as covering the whole journey from search through to after the stay. Maxxton has not said when it arrives or what it will be allowed to do.

The team and the framing

Nino Foudraine leads the work, with a background in econometrics and six years in AI. Gijs Gubbels is on conversational agents, Martin de Smit on backend architecture and Roelant de Looff on frontend.

Foudraine describes the shift as moving from software as a service to what he calls agent as a service, and puts the change on the user rather than the technology. "It's a shift in user behavior," he said. Elsewhere Maxxton makes the sharper point that the useful question is not whether a product uses AI but what it can actually do.

Fair enough, and the company should expect to be held to it.

What to check before you get excited

The figures are Maxxton's own, approximate, and come from a single park. Six weeks to one hour is a striking claim about onboarding, but onboarding time is not the same as competence, and an agent that answers a new starter's questions is not proof that the new starter can handle a difficult guest at 9pm on a Friday. Treat the De Krim numbers as a signal that the thing gets used, which is genuinely the hard part, rather than as a measured productivity gain.

Maxxton has not stated availability, pricing, which product versions include Max, or which languages it handles. For a vendor selling into Dutch, French and German parks, language coverage is not a detail.

The bigger question is accountability. If Max creates a reservation at the wrong rate or moves a booking it should not have, who owns that error and where is it recorded. Permission-awareness answers who could have done it. It does not answer whether there is an audit trail showing that the agent, and not your receptionist, did it. Ask for that in writing before you let an agent write to your live inventory.