Building a Legal Knowledge Base: How One Firm Indexed 10 Years of Case Notes
· 7 min read · Case Study
A Bangalore IP law firm transformed a decade of scattered case notes, memos, and research into a searchable AI-powered knowledge base - and found a trademark argument from 2019 in seconds.
Intellex IP Partners had a knowledge problem. The Bangalore-based intellectual property law firm, founded in 2014, had accumulated ten years of institutional knowledge - case notes, research memos, opposition arguments, hearing transcripts, client advisories, and strategy documents. This knowledge represented the collective expertise of every lawyer who had ever worked at the firm. And it was effectively inaccessible.
The documents lived in a sprawling collection of shared drives, personal folders, email attachments, and in some cases, physical files that had never been digitized. There was no consistent naming convention, no tagging system, and no way to search across the entire corpus. When a lawyer needed to find past research on a specific topic - say, the registrability of colour marks under the Trade Marks Act - they had two options: remember who had worked on a similar matter and ask them directly, or spend hours manually searching through folders hoping to find the relevant memo.
"We had this incredible repository of knowledge built over a decade," says Kavitha Raman, the firm's founding partner. "Every complex opposition we had argued, every research memo we had written, every strategy that had worked or failed - it was all there somewhere. But 'somewhere' is not useful when you need an answer in 30 minutes before a hearing."
The problem was not unique to Intellex. Every law firm that has been in practice for more than a few years accumulates institutional knowledge in unstructured formats. The difference between a firm that leverages this knowledge and one that does not is often the difference between winning and losing - between citing a perfectly on-point argument from a past matter and reinventing the wheel from scratch.
Kavitha decided to address this systematically in late 2025. The firm would build a comprehensive knowledge base by indexing every document they had produced over the past decade. The goal was simple: any lawyer at the firm should be able to search the entire corpus using natural language and find relevant past work within seconds.
The project began with an audit. The firm's IT administrator catalogued every location where documents were stored: three shared network drives, two cloud storage accounts, individual lawyer folders (with permission), and archived email attachments from departed lawyers. The total corpus was approximately 14,000 documents - a mix of Word files, PDFs, scanned documents, and email threads.
The indexing process using Lysa took approximately two weeks. The system ingested documents in batches, extracting text (including OCR for scanned documents), identifying document types, and building a searchable index. For each document, the AI generated metadata: the likely practice area, the legal issues discussed, the parties involved, and key legal provisions referenced. This metadata made the knowledge base searchable not just by keywords but by concepts.
The first real test came three weeks after the knowledge base went live. An associate named Prateek was preparing for a trademark opposition hearing. The opponent had filed a mark that was phonetically similar to their client's registered mark, but visually distinct. Prateek needed to argue that phonetic similarity alone was sufficient for confusion - a point that is well-established in law but requires careful citation of the right precedents and, ideally, examples of how the firm had successfully made this argument before.
Prateek queried the knowledge base: "phonetic similarity sufficient for trademark confusion, opposition arguments." Within seconds, the system returned seven relevant documents. The most valuable was a detailed opposition brief from 2019 - written by a lawyer who had since left the firm - that made exactly this argument with comprehensive citation of Supreme Court and High Court precedents. The brief had been filed in a matter involving pharmaceutical brand names, where phonetic similarity is given particular weight due to the risk of dispensing errors.
"That 2019 brief was gold," Prateek recalls. "It had exactly the argument structure I needed, with citations I would have spent a full day finding on my own. I adapted it to our current facts in about an hour. Without the knowledge base, I would never have found it - I did not even know it existed."
The firm began tracking how often the knowledge base was used and what value it delivered. In the first three months, lawyers conducted an average of 12 searches per day across the firm. Approximately 70% of searches returned at least one highly relevant result. The estimated time savings - based on lawyers' self-reported assessments of how long the alternative research would have taken - averaged 2.5 hours per successful search.
Beyond time savings, the knowledge base produced qualitative improvements in work product. Junior associates could now access the reasoning and research of senior lawyers who had handled similar matters years ago. This effectively compressed the learning curve - a second-year associate could produce work informed by a decade of institutional experience rather than just their own limited exposure.
Kavitha noticed another unexpected benefit: consistency in the firm's legal positions. "Before the knowledge base, different lawyers might take slightly different positions on the same legal question, depending on what research they had done independently. Now, because everyone can see how the firm has argued a point in the past, our positions are more consistent. That matters for credibility - both with the Registry and with courts."
The knowledge base also proved valuable for client pitches and business development. When a potential client approached the firm about a complex geographical indication matter, Kavitha could instantly search the knowledge base and demonstrate that the firm had handled similar issues before - pulling up specific examples of past work (with client details redacted) that showed relevant expertise.
The implementation was not without challenges. Some lawyers were initially reluctant to have their work product indexed and searchable by colleagues - a concern rooted partly in professional pride and partly in worry about being judged for older, less polished work. Kavitha addressed this by framing the knowledge base as a collective asset: "Every memo you wrote, every argument you crafted, every research note you produced - it all has value beyond the matter it was created for. The knowledge base simply makes that value accessible."
The firm also had to address quality control. Not every document in a ten-year archive is equally reliable - legal positions evolve, statutes are amended, and precedents are overruled. The system addresses this partly through date-awareness (newer documents are weighted higher in search results) and partly through a tagging system where senior lawyers can mark documents as superseded or outdated.
For other firms considering a similar project, Kavitha offers practical advice: "Start with what you have. Do not wait until your documents are perfectly organized or consistently named. The AI is remarkably good at extracting meaning from messy, real-world document collections. Upload everything, let the system index it, and then refine over time. The perfect should not be the enemy of the good."
Ten years of knowledge, previously locked in scattered files and departed lawyers' memories, is now available to every member of the firm in seconds. For Intellex IP Partners, the knowledge base has become their most valuable non-human asset - a living repository that grows with every matter they handle and every document they produce.