Document Analysis Intelligence: from many documents to a comparable overview in a table interface

Bouwend Nederland works with large volumes of documentation: party plans, policy papers, reports and memos. The information you need is almost always in there, but you can't find it quickly enough, especially when you want to compare several documents at once on exactly the same questions.

Classic search falls short here: you search for words, not meaning. And a standalone chatbot that retrieves fragments is useful, but stays too ad hoc: you get answers per question, not the overview you need to make decisions.

The opportunity lay in turning documents into a comparable dataset, without having to summarise everything by hand first.

We built a tool that indexes large documents in a vector database, so you can query them in natural language and immediately get the relevant fragments back. With every query the system retrieves the most relevant passages, so you get not only a conclusion but also the evidence straight away: literal pieces of text that support the answer. That keeps you close to the source at all times.

The real power lies not in retrieving a single answer, but in the way questions are asked. That is why we introduced a query matrix: a table in which each row is a topic and each column a focus point. In the rows you define themes such as labour market, nitrogen, safety, housing or innovation. In the columns you set out exactly what you want to know, for example the position, a concrete measure, the substantiation, a chosen category or a score against predefined criteria.

This way you can not only ask "what does this document say about this?", but systematically query "what does every document say about every topic, from every focus point?". The result is a flexible framework that lets you assess dozens of documents at once, in a structured, repeatable and transparent way. You design your question set once and get back an overview that is directly comparable.