White Stork builds knowledge assistants that answer questions from your own documents and data, and show the source for every answer. We turn policies, contracts, manuals, catalogs and records into an AI your team and customers can ask in Arabic or English. It finds the right information first, answers from it, and says so when the answer isn’t there.
Legal research agent
An AI agent that searches ~104,000 laws, ~358,000 versioned articles and ~49,000 court rulings, answering in Arabic with citations to the exact article, court and decision, and using the version of each law in force at the time.
Our own models, reading scanned pages
Lebanon’s laws weren’t available as clean data. They existed as public scanned pages and scattered documents. We trained our own models to read those pages: detecting titles, article text and boundaries, separating each article, and converting the Arabic into searchable text. That pipeline produced the ~358,000 structured articles Co-Lawyer answers from. Most knowledge systems assume clean data already exists. When it doesn’t, we build it.
Six dynamic parts work together, and together they’re why you can trust an answer, not just believe it.
6 parts
Policies, contracts, manuals and records brought together and kept up to date.
Understands how information relates, like which rule replaced which, or which contract covers which client.
Searches by meaning, not exact words, so people get an answer, not a list of files.
If the information isn’t there, it says so instead of guessing.
Uses what’s in force today and ignores outdated or replaced documents.
Works across both languages, including formal Arabic documents.
Built for Lebanese law: over 100,000 laws and 350,000 articles, with every answer cited.
Answers come only from your documents, each one linked, so your team can check it in one click.
Uses the policy, price or rule in force today, not the one it replaced.
Runs on AWS in your own account, delivered by an Advanced Tier Services Partner with the Generative AI Competency.
RAG is a way of building AI that looks up the relevant information in your own documents before it answers, then answers from what it found and shows the source. Answers come from your current information, not from what a general AI model happens to remember.
A knowledge graph stores information as connected facts, not just pages of text: this law amends that one, this contract belongs to that client. That lets the AI answer questions about how things relate, which a keyword search can’t. It’s how Co-Lawyer traces exactly which law is in force.
We make it answer only from your documents, show the source for every point, and say “I couldn’t find that” rather than guess. We also limit it to the topics it’s meant to cover, and test it on real questions from your team before launch.
That’s common, and we handle it. Scanned documents, PDFs and paper archives have to be read and organized before an AI can answer from them, and we build that step ourselves. It’s how we turned scanned Lebanese law data into the ~358,000 structured articles behind Co-Lawyer, using models built for Arabic scans.
Yes. The system runs in your own AWS account, in the region you choose, and follows your existing access permissions, so people only get answers from documents they’re allowed to see. Your documents are not used to train public AI models.