From one search query to a reasoning agent: how vragen.ai finds better answers

For a long time, an answer from your own content only went one way. A visitor asked a question, the system searched your documents once, and an answer came back. For most questions that was enough. But the questions people type into a website have changed. They are longer, more specific and often hold several questions at once. One search query doesn't always do them justice anymore.

This is the story of how vragen.ai grew from that one fixed search query into something that reasons its way to the answer. The promise stays the same: reliable, traceable answers from your own content. How it gets there has improved.

Reliable answers from your own content

A language model on its own only knows what it learned during training. It knows nothing about your products, your policy or the document you published last week. And when it doesn't know something, it fills the gap all too often with something that simply sounds right. That is the last thing you want on your own website.

That is exactly the problem RAG (Retrieval-Augmented Generation) solves. Before an answer appears, the system first looks up the most relevant passages in your own content, and only then writes an answer based on them. The result is an answer you can trust, with the sources it came from. It has been the foundation of vragen.ai from the start, and it is the reason our answers link back to their source instead of asking you to take them on faith.

The limit of one search query

In the first generation of vragen.ai, that searching followed a fixed path. A question came in, the system searched once in a predetermined way, and passed the results to the model. Reliable and fast, and for simple questions it still works fine.

But one limitation was baked in: the search happened only once, and always in the same way. The model, the part that is actually good at language, had no say in what got searched for. It only saw whatever that single query happened to return.

Think about how you look something up yourself. You search, scan the results, notice you phrased it badly and try again. You spot a document you had better read in full instead of one fragment. You narrow it down: only this year, only this category. One fixed search query can't do any of that. If the first attempt misses, the answer suffers and there is no second chance. What really counts is this: can the system search again when the first attempt falls short?

From a fixed path to a thinking agent

So we stopped laying out the path in advance.

Instead of one search query followed by one answer, vragen.ai now works with agents. An agent gets a set of tools and the freedom to decide how to use them. It reasons about the question, chooses whether and what to search for, reads what comes back, sharpens its approach and works through several reasoning steps until it has enough to answer well. We call that agent mode: one question, several reasoning steps, a better answer.

And this matters: it is not a break with what came before. An agent gets a running start. It first runs a familiar search pass to pull in a handful of likely relevant documents, before it even begins to think. The proven foundation we built for classic search hasn't disappeared; it has become one of the reliable tools the agent reaches for. Agents build on our RAG foundation rather than replacing it.

That running start is also why agents are fast. Because this preflight step already retrieves the relevant documents the way our classic pipeline did, an agent answers everyday questions just as quickly as the old setup: it doesn't have to search from scratch. And when a question calls for more, you can let an agent dig deeper and think longer, trading a little speed for a better answer where it matters. By default you get the best of both: as fast as classic search in everyday use, and able to think further the moment a question asks for it.

How an agent tackles a question

Once it starts thinking, an agent moves roughly the way a careful colleague would. It begins with the documents it already has in hand and decides it needs more. It writes its own search term, rewrites it when the first attempt disappoints, and splits a multi-part question into separate searches when that serves the answer better. Where the old path applied one fixed way of processing a question, the agent now makes those choices itself, guided by its instructions. It discovers which filters your content offers: by date, by category, by whatever your documents are tagged with. It applies them to find exactly the right material. When the passage that best matched the question turns out not to hold the answer, it opens the full document and reads it properly. Then it brings everything together into a well-founded answer.

That back and forth is simply impossible on a fixed path. It is the difference between a form letter and a conversation, and it is exactly what makes the difficult, multi-part questions answerable.

Everything stays traceable

A more capable system is only worth having if you can still see what it does. With vragen.ai you can. Every step an agent takes is recorded: every search it runs, the filters it applies, the documents it reads, its reasoning along the way and the sources it eventually cites. When an answer matters, you see exactly how the agent got there.

That is the principle that has guided vragen.ai from the very beginning. No black box, but a platform you can see into and steer, so you never have to wonder whether the AI quietly made something up in your organization's name.

Classic search and agents side by side

	Classic search	Agent mode
Searching	One fixed pass, determined in advance	The agent decides when, what and how, and can search again
Search terms	Set in advance, the same every time	The agent writes, rewrites and splits them itself
Filtering	Configured in advance	Discovered and applied on the spot
Reading documents	The fragments that best match the question	Can open the full document, because the best fragment isn't always where the answer sits
Reasoning	Not about the search	Across several reasoning steps
First attempt misses	Weak query, weak answer	Recovers by searching again

Everything is moving to agents

Agents are becoming the foundation of vragen.ai, and we are moving existing setups across. AI does the heavy lifting of the migration, but we always do it together with you. We look at how your current setup is configured, move it over and check that it behaves the way you expect, before anything changes for your visitors.

The result is a setup that keeps doing exactly what it did. There is nothing to rebuild, and from day one your visitors get the same answers from the same content. What changes is what becomes possible afterward: that same setup can now reason, search again, filter and read whole documents the moment a question calls for it.

The advanced settings you relied on don't disappear either; they move to where the agent can work with them. Things like rewriting a question and splitting it used to be fixed steps in the pipeline. Now they sit in the agent's instructions, which means the agent applies them with judgment instead of the same way every time. You keep the same control over how questions are handled, in a form that adapts to the question in front of it.

What stays the same is the part that always mattered: answers from your own content, with the sources included, and full insight into how they came about.