AI in the real world
Lessons from Morgan Stanley and A&O Shearman
AI in the real world
Lessons from Morgan Stanley and A&O Shearman
Two industry giants shared remarkably candid insights into their multi-year AI transformations, offering a masterclass in moving beyond pilots to production-scale impact. Jeff McMillan (Morgan Stanley) and David Wakeling (A&O Shearman) demonstrated that even the most traditional industries can reinvent themselves with AI.
Morgan Stanley: engineering financial intelligence
McMillan, Morgan Stanley's first Head of AI, revealed the depth of transformation possible when AI is treated as strategic infrastructure. With 20,000 employees building and designing software (25% of the firm), the opportunity and challenge are immense. "Most things we can build ourselves" McMillan stated, challenging conventional wisdom about buy-versus-build. Morgan Stanley has identified eight core AI patterns covering 95% of use cases, from document processing to client intelligence. By centralising development, they've dramatically reduced both cost and regulatory burden.
The results are striking. Research processes that took 40-60 hours now complete in 1-2 hours with higher quality. Document processing that consumed 25% of system time is being automated. Most remarkably, their client intelligence system synthesises meeting recordings, CRM data, and market intelligence to brief executives in real-time.
Key lessons from Morgan Stanley's journey:
- Start with structured evaluation: "9 months, 20,000 pieces of feedback before you're in the high 90% accuracy levels"
- Domain expertise is crucial: "You need your absolutely best people to participate"
- Quality beats speed: "Every data quality problem is a very specific one"
A&O Shearman: reinventing a 500-year-old profession
Wakeling's presentation shattered preconceptions about innovation in traditional industries. As the first law firm to deploy generative AI globally (pre-ChatGPT), A&O Shearman has systematically transformed legal practice. Their Contract Matrix platform exemplifies the power of domain-specific AI. By channelling AI through curated, specialist knowledge - "imagine the AI is trained only on the work of our ten French nuclear energy project finance specialists” - they've achieved productivity gains of "a day a week per lawyer" compared to generic AI tools.
The firm hasn't stopped at internal efficiency. They've created new revenue streams by licensing their AI tools to clients and developing managed AI services. One Morgan Stanley project compressed a 30-day manual review requiring dozens of lawyers into an AI-powered system delivering 75% faster results at 50% of the cost. Cultural transformation proved as important as technology. A&O Shearman's approach included:
- Clear messaging: Using AI is "a basic skill of a lawyer today"
- Celebrating super users and tying adoption to compensation
- Testing AI skills in hiring: "What prompts would you use to answer this legal question?"
- Bold business model shifts: Moving from billable hours to fixed fees while preserving margins
"The first one is, we focus on very clear and simple messaging that our lawyers should be using AI, and more so they should see the use of AI as a basic skill of a lawyer today. We reinforced that by promoting the profile of our super users internally, those attorneys around the world who we could see on the usage stats were really using it."
Common success patterns
Both organisations emphasised similar critical factors:
Human-in-the-loop remains essential:
"Everything we use is human-enabled. We don't have autonomous vehicles yet," noted McMillan. Quality comes from combining AI efficiency with human judgment.
Reuse and patterns drive scale:
Rather than bespoke solutions, both firms identified reusable patterns. McMillan: "99% of what we do we're able to do through prompting."
Cultural change requires top-down commitment and bottom-up adoption:
Both organisations invested heavily in education, clear expectations, and making AI adoption mandatory rather than optional.
The future is collaborative:
McMillan predicts cross-industry convergence: "They're going to start selling financial services products. And we're going to start selling legal advice."
Looking forward
Both presentations revealed a crucial truth: AI transformation in regulated, traditional industries is not only possible but accelerating. Success requires patience, systematic approaches to quality, and willingness to fundamentally reimagine business models.
As Wakeling summarised: "Our objective is to shift in suitable areas from a billable hour model to a fixed fee model, make it cheaper for our clients, and preserve or probably increase our margin." This isn't just efficiency - it's reinvention.
