Rukkor AI

Phase 1: Document and file analysis

The first service we will release is the ability to “talk” to your documents and files. After uploading your files, you can choose to enable “available for AI” for each file. The document is then analyzed and added to the underlying knowledge of the AI agent. This makes it possible, for example, to upload a room specification and then interact with it by asking questions such as: “In which rooms is there parquet flooring?” or “How many types of a certain material are present in the project?”

This step is based on a text model that enables analysis of text and provides answers to user questions, similar to what you see in ChatGPT, Copilot, and Le Chat.

The benefit of this type of interaction with your project documents is that it becomes easier to find information, create summaries, or check whether something is included or not.

Phase 2: AI-based measurement of drawings

Being able to identify, for example, rooms on a drawing and mark them as is currently done manually is part of phase two of our development. Our current models are already showing good results, enabling reliable measurement of areas.

We are now building training data and training models to identify elements in drawings and place objects that can be further worked with.

The potential benefit of AI-based measurement is significant but also comes with certain caveats. Blindly trusting AI measurements can have serious consequences. Therefore, the credibility of results is at the top of our priority list. As a user, you must be able to trust the results, so the work does not simply shift from manual measuring to checking every object the AI generates.

This step is based on a visual model that we are continuously training to improve its ability to identify correct areas, walls, doors, and other details.

Phase 3: Information completion from document analysis and AI-based measurement

The third phase of the project aims to combine document analysis (step 1) and AI-based measurement (step 2) to propose calculation rows based on the content found in documents and drawings, as well as apply any templates created in Geometra.

This means that you first upload your specifications and make them available for the AI to analyze. Then you upload your drawings and use AI-based measurement to place objects on them. Finally, you use information completion to let the AI suggest rows for the objects created, based on the available documents and the object’s naming.

At every stage, you as a user will clearly see which content has been suggested by the AI and decide whether to use it or not.

Our goal with these steps is to gradually increase the possibilities for you as a user to integrate AI into your work—without compromising the reliability of the results. Without credibility, the point of using AI in your workflow is lost.

Development funded with EU support

In July, we received good news from Almi and Region Skåne. The presentation we made to them about our vision for implementing AI was very well received, and we were granted public funding for further development of our AI-based services.

This support makes it easier for us to carry out our initiative and accelerate the development we have been working on for a long time. We are, of course, very pleased with this and would like to extend a big thank you to Almi and Region Skåne for their trust in us and for the public funds we have been granted.