The vragen.ai roadmap for 2026: from answers to control over conversations

At SWIS we see every day how AI is becoming a fixed part of how organizations help visitors, guide them and prompt them to act.

So this year we are developing vragen.ai further, around two closely connected themes: the Conversation Management System and self-service.

The shift: from RAG to a Conversation Management System

We have written before about the change from traditional websites to a more conversational and generative kind of interaction (see the article on Frankwatching, in Dutch, and our earlier piece on agents, MCPs and widgets).

Users don't come to search, they come to achieve something. Which means the systems we use have to change too. A stand-alone chatbot or search function is no longer enough. What is needed is a system that understands intent, combines information, carries out actions and keeps an eye on quality and results at the same time. That is what the concept of a Conversation Management System is about.

Our roadmap works toward such a system step by step.

Theme 1: the Conversation Management System

Within this theme we strengthen the foundation that makes conversations with AI manageable and scalable.

The knowledge layer as the foundation

Everything in vragen.ai starts with the knowledge layer. Where traditional systems often treat content as separate pages, here it is about coherence. The system has to understand where information comes from, how sources connect and in what context something is relevant.

Besides text, we also look at images, videos and documents. Information changes constantly, and the system has to keep up with those changes without becoming less reliable.

So we focus on organizing and using the information that already exists intelligently.

Brand identity and governance

Now that AI plays a larger role in how organizations communicate, that communication has to stay consistent and controlled. After all, the AI speaks on behalf of the organization.

So we are working on a central place where organizations set down their tone of voice, and the rules and goals that steer how the AI behaves. That keeps the interaction in line with the organization's identity and positioning.

That includes managing escalations. There are always situations where it is better for a person to take over the conversation, for instance with complex or sensitive questions, or when the AI is uncertain. The system has to make that handover smooth, with the full context preserved.

From text to interaction

Conversational interfaces consisted mainly of text for a long time. Now we see a clear shift toward more visual and interactive forms. Within vragen.ai, answers are increasingly supplemented with interactive elements such as buttons, forms and choices. That changes the role of AI. The system gives information and helps users take a next step right away.

Users make choices or enter details inside the conversation itself, without going to a separate page. Information and action come together in one interface.

Insight as a way to steer

As conversations with AI become more important, organizations above all want to know what those conversations deliver.

Within vragen.ai we are working on analytics that show this from several angles. Organizations see which conversations contribute to goals such as applications or conversions. They also see what users want and what they need. And we look at quality: how well does the AI help, and where could it do better?

From understanding to doing

The next step for such a system is the one from understanding to doing. That takes skills and flows: reusable building blocks that determine how the AI acts on an intent.

In practice the system recognizes what someone wants and knows which steps are needed to reach that goal. Sometimes that is a simple flow, sometimes a more complex process involving several systems.

Through a connection with a CRM, for example, the AI can also carry out actions inside existing processes. The AI's role shifts from advising to supporting and executing.

Testing and manageability

The more people set up and improve a conversational system, the clearer it has to be who can change what, and how changes go live.

So we are investing in features that let organizations test and check new settings, content or flows before they go live.

Personalization as the next step

A conversational interface only really helps users when it takes their context into account. Think of a logged-in user, earlier interactions or personal preferences.

Within vragen.ai this is already partly possible through API and MCP connections. During a conversation the system can retrieve information from other systems in real time. That lets the AI adjust answers based on an account or a role, for instance.

As we develop further toward a complete Conversation Management System, we are exploring how personalization can play a bigger part.

Theme 2: self-service

Besides adding new capabilities, this year we are making the product easier to use.

vragen.ai has been around for three years now. In that time we have extended the platform and let it grow along with developments in AI. At the same time we see that managing it isn't always easy for organizations. There is a lot you can do, but it isn't always immediately clear where and how to change something. We want to change that.

More control for the user

The goal is for organizations to manage and improve their own conversational environment. You adjust things yourself, without needing our team for changes.

For that, the system has to be clear and easy to understand.

A new admin environment

Together with our designers we have redesigned the admin environment. Features are grouped more logically and you can see more clearly what you can adjust.

The focus is on clarity and simplicity. You can do as much as before, but you see more clearly how settings connect.

Support for optimization

We are also making it easier to keep improving your environment. Conversations deliver valuable insight, but that insight has to be turned into concrete action.

So we are working on features that help organizations quickly see where improvement is needed and how to make it happen.

Accessibility and multiple languages

In its conversations with visitors, vragen.ai is already multilingual. A visitor can ask a question in their own language and gets an answer in that language, even when the underlying website or knowledge base only holds Dutch content. So information stays accessible to a wider group of users, without an organization having to translate all its source content.

What is currently Dutch-only is the admin environment where organizations configure and improve vragen.ai. The multilingual work is therefore about that admin environment: we see more and more teams wanting to use it in other languages.

A roadmap as a direction, not an end point

Developing toward a Conversation Management System isn't an end point you reach in one go.

This roadmap takes steps in that direction.

Our team keeps working on vragen.ai in the months ahead, learning from what happens in practice.

Missing something on this roadmap, or running into something? Let us know through the contact form. We build vragen.ai together with customers, so your feedback helps decide what we pick up next.