Bringing together vast volumes of financial data spread across multiple systems is a persistent challenge for financial institutions. At LSEG, this meant managing a fragmented data estate comprising petabytes of financial data across 30 systems and 1,200 datasets, driving its partnership with Microsoft to modernise data infrastructure and build a more unified, AI-ready foundation for next-generation data and analytics solutions.
“When you need to pull data together across disparate sources that are in different formats and varying levels of modernisation and maturity, it makes it difficult to react to market demand quickly,” says Dave Byrne, group head of data platforms at LSEG. “We knew it would be far more efficient to bring everything into a single, modern platform. It would mean we could run the organisation leaner and react to market demand faster.”
The decision to partner with Microsoft to achieve that consolidation was, says Byrne, as much about people as technology. “The organisation and the people you are partnering with are just as important – probably more important – than the technology you pick,” he explains. “Our decision was about finding the right collaborative partner to co-engineer solutions to business problems.”
The result has been a 10-year partnership that goes far beyond a typical vendor relationship. Microsoft’s executive vice president Scott Guthrie sits on the LSEG board and hundreds of engineers from Microsoft’s commercial engineering organisation work in blended teams alongside their LSEG counterparts.
“When we set out to build something, it is never simply a case of Microsoft launching a product,” says Marc Esmiley, head of product at Microsoft’s Financial Services Studio. “The starting point is always: how can Microsoft technology support the evolution of what LSEG wants to deliver to its end users? It is a co-innovation partnership, not a traditional vendor relationship where Microsoft provides the technology and LSEG takes it from there.”
For Byrne, the value shows up most clearly in joint product roadmap sessions. “Some of the most enjoyable things I get to do in my job are when I travel out to Microsoft and we discuss product roadmaps at a very early phase, where it is still early enough that we can provide meaningful input. I have not had this kind of relationship with any other partner.”
Some of the central strategic questions the partnership has had to answer are deceptively simple: when should LSEG data be embedded into existing tools, when should existing surfaces be enhanced and when does something entirely new need to be built? The answer is to start from where the customer already is. “We want to embed data into Microsoft Teams, Copilot Studio, Foundry, Claude, ChatGPT – giving users the same capabilities and experience they are familiar with and letting them choose their preferred endpoint,” says Esmiley. “We are not saying to customers you need to change your whole infrastructure to adopt this new application. This reflects a broader shift in how organisations are redesigning operating models for the AI era.”
Byrne frames it as a philosophy that predates AI but has been dramatically accelerated by it. “In a pre-AI world I would have said it was important for LSEG to meet customers wherever they want to consume our data,” he says. “AI has only accelerated and expanded that problem statement.”
Centralising data in Microsoft Fabric before distributing it flexibly has become the architectural answer to a world where, as Byrne puts it, “nobody knows what the flavour of the month will be a year from now.” Crucially, though, distribution alone is not enough. “It is not just about getting data to where clients want it but delivering it in a format that is ergonomic for native AI capabilities: business glossary definitions, data model semantic definitions, responsive APIs and model context protocol (MCP) servers.” This approach aligns with emerging best practices for building AI-ready operating models and integrating data into modern workflows.
Before any of that AI-era distribution could be contemplated, however, both organisations needed to build governance foundations capable of bearing the weight of financial services regulation. Data lineage – the ability to trace every data point from source to destination – emerged as a foundational priority from the earliest days of the partnership. “It is not at a table-to-table level, not even column-to-column, but data cell to data cell,” says Byrne.
Working in the financial services sector also places uniquely demanding requirements on historical traceability. “It is not enough to know what the right value is – you have to know what your understanding of the historical value was at a given point in time, including when and why any corrections were made,” says Byrne.
Esmiley credits LSEG with pushing Microsoft on this explainability. “Our first agent launch would provide an answer, and LSEG said: ‘We need to expose that lineage to the business user’.” Making an agent’s reasoning transparent required significant infrastructure investment. “If the answer is six, the user needs to understand how it arrived at six and what data it relied on,” explains Esmiley. “User disagreement has become a valuable signal in its own right – a powerful feedback loop for LSEG’s engineering roadmap.”
The partnership’s most recent milestone is the launch of a model context protocol (MCP) server enabling customers to build AI agents directly on LSEG data. Byrne sees three distinct client archetypes emerging: those who want raw data and nothing else; those who want a fully turnkey AI experience; and a growing middle ground for whom MCP represents an appealing option.
“They do not want to build something completely custom from the ground up, but they want to run their own large language model (LLM) and interact with their data in the most seamless way possible,” Byrne says. “I think that hits a real sweet spot.”
By both executives’ accounts, this is only the beginning. Byrne and Esmiley acknowledge that the AI phase of their partnership arrived far faster than they anticipated. “We agreed on all this amazing work as part of a 10-year partnership, and then roughly three weeks after the official announcement in December 2022, AI models really started to connect commercially with the market,” says Esmiley.
Since then, the relationship has evolved in real time, with both organisations continually rethinking how they collaborate day to day. Yet for Byrne, the pace of change only reinforces how early the industry still is in the journey ahead. “If I were comparing it to the personal computing revolution, we are not even at the Apple II phase yet,” he says. “We are only beginning to see what’s possible. It’s incredibly exciting.”
Discover more insights from Microsoft Partners in the Summer 2026 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox.