London Stock Exchange's AI Transformation: How 33PB of Data Powers Intelligent Financial Analysis

Introduction

In financial services, AI adoption is moving from the experimental phase to scaled deployment. The London Stock Exchange Group (LSEG), a global financial market infrastructure and data provider serving 44,000 clients across 170 markets, has been at the forefront of this shift. Emily Prince, LSEG's Group Head of Analytics and AI, shared the company's AI transformation journey at an OpenAI event, revealing how a centuries-old financial institution is finding its balance amid rapidly evolving technology.

LSEG's history dates back to 1698, making it one of the world's oldest stock exchanges. In 2021, LSEG completed its $27 billion acquisition of Refinitiv — one of the largest deals in the financial data industry's history — transforming LSEG from a traditional exchange operator into a comprehensive financial data and analytics provider, putting it in direct competition with the Bloomberg Terminal. Refinitiv was formerly Thomson Reuters' Financial & Risk business division, home to iconic products like the Eikon terminal and Datastream. The acquisition brought LSEG massive financial data assets but also enormous data integration challenges — and that's precisely where its AI strategy begins.

LSEG Everywhere: The Core Philosophy of the AI Strategy

From Data Silos to a Unified Intelligence Platform

LSEG's AI strategy is called "LSEG Everywhere," built on a core principle: no matter where users work or what tools they use, they should be able to anchor their work on trusted financial information. This isn't just about providing generic summaries of information — it's about fundamentally transforming how people work.

Emily Prince noted that LSEG has gone through multiple acquisitions, accumulating a wide variety of data models and systems. The data and analytics business alone holds over 33PB of data. To put that in perspective, the entire digitized collection of the U.S. Library of Congress is approximately 15PB. LSEG's data spans real-time market quotes, historical trading records, corporate financial statements, ESG ratings, news, alternative data, and more, sourced from 170 markets worldwide and covering asset classes including equities, bonds, foreign exchange, commodities, and derivatives. The data's time horizon ranges from decades of historical records to millisecond-level real-time pricing, forming the foundational information layer for financial decision-making. How to unify these scattered data assets and serve both internal teams and external clients in a scalable way is the core challenge they face.

MCP Protocol: A Key Step in Unlocking Financial Data

LSEG partnered with OpenAI to achieve unified data access through the Model Context Protocol (MCP). MCP is an open-source protocol standard released by Anthropic in late 2024, designed to solve the connectivity problem between large language models and external data sources. Before MCP, every AI application needed custom integration solutions to connect with different data sources, resulting in massive duplication of engineering effort. MCP provides a standardized "universal interface," similar to what USB-C does for hardware devices — through MCP, AI models can securely query databases, call APIs, and access file systems without writing specialized adapter code for each data source. In 2025, OpenAI integrated MCP support into ChatGPT, enabling third-party data providers to embed their data services directly into ChatGPT's conversational interface.

This means users can ask specific financial questions and generate reports directly within ChatGPT, with answers grounded in LSEG's trusted data rather than generic information.

More importantly, this integration is "plug and play" — no six-month data onboarding process, no massive project engineering. Users can generate analysis directly in spreadsheets that previously would have taken hours to complete. Emily admitted that this efficiency gain is "both exciting and a little depressing" — because it makes you realize how much time was previously spent on inefficient work.

From Experimentation to Scale: Building an AI Evaluation Framework

Strategic Convergence After a Thousand Flowers Bloom

Emily described LSEG's AI evolution as going "from a thousand flowers blooming to scalable impact." During the early experimentation phase, numerous independent AI projects emerged across the organization. But achieving true scale requires building a systematic evaluation framework.

She emphasized that "evaluation framework" is one of the phrases she uses most frequently every day. This framework needs to answer several key questions:

• What problem are we solving?
• What does success look like for different teams (finance, marketing, product, engineering, client-side IBD, portfolio management, research analysis)?
• How do we maintain quality standards across all use cases?

It's worth elaborating on the specific work contexts of IBD and portfolio management. IBD (Investment Banking Division) is the core department within investment banks responsible for M&A advisory, IPO underwriting, debt financing, and other services. Analysts' daily work includes creating pitch books, building financial models, conducting comparable company analysis, and precedent transaction analysis — all of which are highly dependent on data access and document generation. Portfolio management involves asset allocation, risk monitoring, performance attribution, and other processes that require continuously tracking the performance of hundreds or even thousands of securities. Both roles face severe information overload, making them prime candidates for AI augmentation.

Systems Thinking Beyond Model Selection

You might not have noticed, but LSEG's evaluation system goes beyond simple model selection. They focus on success criteria for different use cases and how to maintain consistent quality levels during scaled deployment. The universality of their API strategy and flexibility in model selection allow them to iterate quickly without being locked into a single technology path.

AI Transformation of the Financial Analyst Workflow

Core Pain Point: Information Overload and Time Constraints

For financial analysts, the core problem AI solves is the tension between breadth and depth of information access. Ideally, analysts want access to all relevant information, especially those "slightly orthogonal" insights — which often provide differentiated analytical perspectives and competitive advantages.

But the reality is that everyone's time is limited. For a long time, analysts have had to condition themselves to pull data from only specific information sources. AI changes this dynamic: analysts can extract information from richer sources while maintaining quality standards through built-in criteria, policies, and preference settings.

From Hours-Long to Real-Time Analysis Iterations

LSEG was among the first financial institutions to launch applications within ChatGPT, creating new interaction interfaces for analysts and clients. Analysis iteration cycles that previously took hours or even days can now be completed in real-time conversations. This isn't just an efficiency improvement — it's a fundamental shift in how work gets done.

This "workflow compression" phenomenon is happening broadly across the financial industry. JPMorgan Chase CEO Jamie Dimon has said AI could shorten the workweek to 3.5 days. Goldman Sachs research estimates that generative AI could affect approximately 35% of tasks in financial services. Specifically, writing an equity research report used to take an analyst team about a week, involving data collection, model updates, peer comparisons, and report drafting. With AI, data collection and initial draft generation can be completed in minutes, shifting the analyst's role from "information mover" to "insight validator" and "judgment maker."

Governance and Culture: The Two Pillars of AI at Scale

Responsible AI Principles: Empowerment, Not Constraint

LSEG began systematically building its responsible AI principles and governance framework two years ago. Emily emphasized that their goal is not to "put handcuffs on people" but to create safe scaffolding within which teams can innovate freely.

This governance framework must be understood within the special regulatory context of the financial industry. In the EU, the AI Act classifies financial AI applications such as credit scoring and insurance pricing as "high-risk" systems, requiring conformity assessments and human oversight. The U.S. SEC and FINRA are also intensifying regulatory scrutiny of AI-driven investment advice. Additionally, financial institutions must comply with data privacy regulations (such as GDPR), model risk management guidelines (such as the Federal Reserve's SR 11-7), and anti-money laundering (AML) compliance requirements. This means financial institutions cannot simply copy tech companies' approaches to AI governance — regulatory compliance must be a core design principle.

The specific approach: rather than creating an entirely new set of rules, they examine end-to-end build and service processes, identify areas that need adaptation, and embed governance into daily workflows. As workflows compress (from ten-person teams down to one or two people solving problems in less time), ensuring that governance and processes remain embedded is an ongoing, evolving challenge.

Cultural Transformation: From Fear to Active Engagement

In Emily's view, AI adoption is more of a cultural issue than a skills issue. She observed that the people who truly "take the leap" share common traits: an open mindset and a spirit of active experimentation. But alongside excitement, there is often fear and misunderstanding about AI.

LSEG's response strategy operates on three levels:

1. Empowerment: Ensuring everyone has access to tools like ChatGPT and MCP
2. Education: Providing systematic learning programs
3. Practice: Shifting from awareness-building to hands-on building, solving specific rather than generic problems

Implications for the Financial Industry

An Era of Self-Disruption Without a Blueprint

In her closing remarks, Emily offered a profound insight: now is the best time for the financial industry to "self-disrupt." Many of the cumbersome processes and work methods accumulated over the past 20 years weren't designed intentionally — they're the result of historical evolution. Now is the time to boldly ask: does this process truly have a regulatory requirement to be this way? Or can we reinvent it?

She acknowledged that there is no ready-made blueprint for this era — no book, no precedent to reference. Everyone is learning and creating simultaneously. This uncertainty itself is precisely where the opportunity lies.

Conclusion

LSEG's case illustrates the typical AI transformation path for large financial institutions: from strategic positioning (LSEG Everywhere) to technical infrastructure (MCP and API strategy), from evaluation frameworks to governance systems, from cultural change to continuous iteration. In this era "without a blueprint," maintaining an open mindset, building flexible frameworks, and iterating quickly through validation may be the most pragmatic way forward.

Key Takeaways