UBS enhances in-house legal search using Azure AI Search and Azure OpenAI Service

UBS’s in-house legal team must find very specific information, like a clause or regulation, across a library of 26 million legal documents in multiple languages. The existing search system was inefficient, and UBS wanted a more effective solution.

Working with Microsoft Azure OpenAI Service and Azure AI Search, UBS built the Legal AI Assistant (LAIA) to help employees pinpoint phrases, clauses, and paragraphs using natural language and semantic similarity, rather than keyword matching.

UBS employees can now locate information much more quickly and easily than they had with the prior search tools produced. Employees are finding productivity gains, thanks to the improved search experience and accuracy provided by LAIA.

Navigating an ocean of legal information

To best support their sector-leading global wealth management enterprise, UBS’s in-house lawyers and legal experts needed to effectively search legal documents across multiple languages.

“Our internal document library contains about 26 million documents,” explains Vlad Stoian, Product Owner for the Legal AI Assistant at UBS. “Finding specific knowledge in this vast repository was like finding a particular grain of sand on the beach.”

The search tools in place weren’t up to the task. “It took lots of manual work to do things like searching provisions across funds,” says Felicia Efta, Legal Counsel Asset Management, Head of Alternative Funds EMEA at UBS.

Improving legal workflows with Azure AI knowledge retrieval

UBS sought to empower its legal team to focus on more value-added work and streamline the time spent searching through documents. The company chose to build its solution on Microsoft Azure OpenAI Service and Azure AI Search. The resulting knowledge management solution, Legal AI Assistant (LAIA), offers semantic search functionality at scale using metadata identification from the underlying storage systems.

“The AI assistant vectorizes language in existing documents, grouping related concepts together,” explains Stoian. “It can understand natural language and pick up on synonyms and semantic similarities. For instance, if someone searches ‘remaining part,’ LAIA can find ‘legacy portfolio.’ It lets employees enter descriptive text, rather than just keywords.”

LAIA points employees to relevant paragraphs from millions of documents, highlighting key information while maintaining appropriate document-level security protections. “LAIA respects access rights by ensuring people only search and access documents they’re authorized to view,” says Stoian. “Microsoft technology has helped us respect document-level access in a consistent way, which was essential given our large volume of files.”

“At UBS, we were early adopters of Azure AI technologies, and they’ve only gotten better over time.”

Ilias Fotopoulos, Lead AI Engineer, UBS

UBS is expanding the tool globally, all while respecting the tight regulations placed on the industry, along with the needs of its employees. “Moving to a system like this takes training,” says Efta. “We need to switch from keyword search to descriptive search and learn prompting skills. That’s why we’ve also developed bespoke training focused on helping employees to improve their queries and better utilize LAIA.”

Microsoft supported UBS with technical guidance to ensure success throughout the development and deployment process. “We did a lot of upfront planning to ensure the solution would fit within the highly regulated UBS environment and work for our use case,” says Ilias Fotopoulos, Chapter Lead, AI & Data Engineer at UBS.

Fotopoulos was impressed by how Microsoft technology facilitated the solution UBS needed. “At UBS, we were early adopters of Azure AI technologies, and they’ve only gotten better over time,” he says. “They’ve helped us streamline search across millions of documents at scale and with speed.”

“Deploying Microsoft AI technologies has made our legal teams’ work easier and faster.”

Vlad Stoian, Product Owner for Legal AI, UBS

Transforming information access and delivering AI-powered search

With the implementation of the Legal AI Assistant, UBS has transformed its knowledge retrieval processes, and lawyers can easily access stored information in a fraction of the time it used to take. UBS legal experts can now conduct complex queries within minutes, not hours, across the vast document repository. This efficiency is freeing up employees to focus on higher-value tasks.

“Deploying Microsoft AI technologies has made our legal teams’ work easier and faster,” says Stoian. “We plan to enhance LAIA further by converting unstructured data to structured data to make it easier to search and filter documents based on specific criteria. We want to keep unlocking more use cases and improve LAIA’s overall utility.”

“The performance of Azure AI Search exceeded our initial prototype’s results, and it’s helped us build one of the largest-scale knowledge retrieval systems at UBS,” says Fotopoulos. “That just reinforces our certainty that collaborating with Microsoft on this project was the right decision.”

Discover more about UBS on Facebook, Instagram, LinkedIn, and YouTube.