Recordings2 October 20253:02
Contract analysis: clause risk, scored and sourced
A service agreement uploaded, read and split into clauses; each clause analysed for risk, given revised wording, and linked to the case law and regulation it rests on.
What it shows
A service agreement goes in. The pipeline extracts the PDF, classifies the document, pulls out the metadata (parties, dates, type) and then the clauses. Pick a clause and the analysis streams in: a summary, the issues found, the proposed changes, a revised clause ready to use, and the rationale behind it.
Inside the analysis, every case, statute or regulation the system relied on is a link; one click opens its full text. The lawyer checks the call against the original rather than trusting the model's reading of it. Recorded by the BioSeek team in October 2025.
The system
Behind it is the US legal knowledge graph from chapter VI: federal and state law, amendments and court audio with speaker identification, queried by agents that are forbidden to use their own legal memory. What they cite, they retrieved.
- Recorded2 October 2025
- Length3:02
- ChapterChapter VI · The cartographer, 2018 – today
- The systemUS legal knowledge graph, 2024
- Live todaybioseek.eu
- LoomWatch the original →
Transcript
Hi everyone. Today I'll walk you through the document section of our app. We'll start with uploading a file, follow its processing pipeline, and finish with a deep dive into clause analysis. Let's begin by uploading a document. Here, I have a service agreement. Once it's uploaded, I click the process file button to start.
As soon as processing starts, you'll see the progress step by step: extracting the PDF content, classifying the document, extracting metadata, and finally, extracting the contract clauses. Each step runs automatically, and when the pipeline completes, the document is ready to use. After processing is done, we navigate to the document page.
Here we have three main tabs. Info. This shows all metadata about the contract, parties, dates, type, and more. Document. Here you'll find the full extracted content of the document. Clause analysis. This is where the real power comes in. In the clause analysis tab, you see a list of all extracted clauses on the left. When I click on one, the panel on the right opens with a button to analyze this specific clause. Let's try it.
Once I click analyze clause, you'll see a progress bar as the system works. At a certain point, the results begin to stream live, so you can follow the analysis in real time. When the analysis finishes, the system takes a little more time to generate improvement suggestions. Here's what that looks like. At the top, you get a clear summary of the clause. Below, the issues identified are listed, showing exactly where the risks or problems lie. Then, the system provides proposed changes to make the clause stronger and more balanced. You also see a revised clause example that's ready to use. Finally, the rationale explains why these changes matter, so you can understand the legal reasoning behind the suggestion. And if you're happy with the recommendation, you can accept it right here.
And here's something really useful. Inside the analysis text, the system automatically highlights data entities it has retrieved, like cases, laws, or regulations. These appear as clickable links. When I click one, a modal opens, showing me the full text of that entity. This way I don't just see the AI's interpretation, I can immediately check the original source in context.
So that's the full journey, from uploading a contract, running it through our processing pipeline, and diving deep into specific clauses. It's designed to give you clarity, save time, and make working with complex documents effortless. Thanks for watching.