Microsoft CELA cuts regulatory compliance review time with GenAI-powered assistant

Microsoft's Corporate, External, and Legal Affairs (CELA) team operates across global jurisdictions, where the pace of regulatory change creates a persistent operational burden. Legal professionals must track, interpret, and act on a continuous stream of regulatory updates — a task that grows more complex as the volume of regulatory content increases across markets. Manual review processes struggled to scale with this demand, consuming significant attorney time on comprehension work rather than higher-value legal analysis. The inability to systematically process regulatory change efficiently posed real compliance risk and limited the legal department's capacity to serve the broader organization.

EY Law and technology teams partnered with Microsoft CELA to build a GenAI-powered conversational assistant integrated with the EY Global Tax Platform. The tool is powered by Microsoft Azure OpenAI and Microsoft Semantic Kernel, built through a phased approach: an intensive two-day workshop addressing data strategy, privacy, and legal requirements; a joint hackathon producing a proof-of-concept; and a workflow POC refined with CELA's regulatory and product teams. The POC automated regulatory comprehension tasks, enhanced compliance monitoring, and streamlined document creation workflows. Within weeks, the EY technology team developed the required automation skillsets, clearing the path for a full pilot deployment within Microsoft's legal department.

The GenAI solution significantly reduced the time legal teams spent comprehending required regulatory compliance actions — the primary bottleneck in the existing workflow. The pilot implementation moved forward following strong hackathon results and positive user feedback from CELA attorneys.

• Reduced regulatory review time: Comprehension of complex regulatory requirements accelerated materially versus manual processes
• Successful POC-to-pilot progression: Compelling assistant responses during the hackathon directly justified advancing to a formal pilot
• Replicable framework: The collaboration established a transferable methodology for AI adoption in legal operations that other organizations can follow

• Start with available data: A small, focused dataset is sufficient to begin building meaningful AI use cases — legal teams should not wait for perfect data conditions.
• Assemble cross-functional teams early: Combining legal, technology, AI policy, and business expertise from the outset accelerates both design quality and stakeholder buy-in.
• Use hackathons to de-risk adoption: A joint hackathon produced a working POC and cleared internal skepticism before any significant deployment commitment.
• Build for the end user from day one: Incorporating CELA practitioner feedback during the workshop phase ensured the tool addressed real workflow pain points rather than assumed ones.