Platform Pulse: AI agents are changing online platforms

A new user of platforms

Many platforms are built for people. But that is changing fast. People search, compare, and make choices through a user interface. More and more, AI-tools are taking over part of that interaction. Think of AI-assistants that compare products, plan trips, or search for vacancies. Not just by showing information, but also by independently carrying out actions.

During the session, Wout and Rens discussed what this means for platform companies. An agent wants to retrieve data directly, compare options, and carry out tasks. That requires different choices in how your platform works.

Trust becomes more important

AI makes processes faster and smarter. But it also raises a new question: how do you keep control when AI makes independent choices?

During the session it became clear that this is not just a question about the behavior of AI, but mainly about how you set up your platform. That's where the MCP-server came back in.

An MCP-server sits as a layer above your platform and describes which actions are available for AI-agents. You can see it as a menu card for AI. Where a user makes choices through an interface, an AI-agent sees directly through the MCP-server which actions are possible. For example, retrieving data, carrying out actions, or starting processes. That also means that you as a platform explicitly determine what an AI can and cannot do.

With that, trust becomes less of something abstract. It shifts to the foundation of your platform: which actions do you make available, how are they limited, and how predictable is the system that responds to them?

Therefore, you can conclude that the reliability of AI is not just in the output, but specifically in that underlying layer of control and structure. Especially with sensitive processes, human control remains necessary, but the MCP-layer increasingly determines how that control is technically enforced. AI can take a lot of work off your hands, but not every decision has to be fully automated. By giving people a role in the final assessment or approval, you keep a grip on the outcome and the process remains transparent.

From interface to infrastructure

AI-agents work differently than people. They communicate easier through APIs, structured data, and protocols. That is why the technical foundation of a platform is becoming increasingly important and the user interface less important.

That doesn't mean that websites as we know them today will disappear. But platforms will soon also have to work for systems that act on behalf of users.

Start small, but start now

A clear conclusion of the afternoon: many organizations can experiment faster than they think. Many platforms already have the necessary data, APIs, and technical connections. This makes it relatively easy to test AI-applications within existing platforms. Think of:

• conversational search

• personalized recommendations

• AI-support within support

• smart matching or filtering

By experimenting, you can discover in a low-threshold way where AI really adds value for your users and your organization. And scale up as soon as a formula proves successful.

To do this, teams do need to collaborate differently. Marketing, product, and development must pursue the same goal, switch faster, and learn together what works.

Platform Pulse: building the future of platform business

With Platform Pulse we create space to discuss knowledge, experiments, and developments around digital platforms with each other. Based on practical experience, experiments, and honest conversations about what works and what doesn't work yet.

The first edition shows how much need there is for that conversation. About technology, but especially about the choices that platform companies must make now.

Curious about the impact of developments like AI-agents, platform business, and digital innovation? Through our newsletter we regularly share insights, practical experiences, and new developments from the platform world. Sign up!