By Alice Chambers |
Not long ago, creating an AI agent meant securing budget, coordinating with engineering teams and waiting weeks – sometimes months – to see an idea take shape. Now, financial services professionals can use Microsoft Copilot Studio to build and refine their own agents in minutes with access to 33 petabytes of LSEG-licenced financial data delivered through an LSEG managed Model Context Protocol (MCP) server. This self-service approach represents a step change in speed and efficiency.
“We’re making our trusted LSEG data available as part of Copilot Studio so customers can build custom agents,” says Emily Prince, group head of analytics and AI at LSEG. “When a customer builds their own agent, they can also blend it with their own proprietary data, which is very powerful.”
This flexibility means firms are no longer constrained to single-purpose tools or predetermined use cases. Instead, agents can be orchestrated to work together, each performing a specific role as part of a wider workflow.
“Agents don’t have to be limited to any one field,” says Prince. “There are all sorts of different ways we can build and organise them from credit agents, to signal agents that capture market events and turn those into action, or even customer-support agents.”
A major factor is the democratisation of agent building itself. Less than six months ago, developing even the simplest agent required highly technical skills or direct support from engineering teams, which slowed down experimentation and increased the cost of innovation. The availability of LSEG-licensed data within Copilot Studio changes that.
“Copilot Studio is a democratised platform where people can build agents without needing engineering help, which is very liberating,” says Prince. “Teams can now self-serve and build entire applications without a single piece of help.”
That shift is most visible in the time it takes to get started. What once required a formal project plan can now be done in the space of a meeting.
“Agents can be very fast to build – it could be completed within a couple of minutes,” says Prince. “Of course, complexity adds development time, particularly when users focus on validation and quality assurance. But even that work is more accessible. Mechanically, it is not a complicated thing to do – it’s very controllable for the end user.”
Financial services teams are already experimenting with a range of scenarios. A credit analyst, for example, might create an agent that brings together market signals, validated data sources, internal documentation and automated alerts.
“You might specify the intent and the descriptor of what the agent is trying to achieve, set the knowledge sources and trigger an email to the portfolio management team based on a market event,” explains Prince.
What makes this especially valuable is the ability to surface the output in whichever tool the end user prefers. Whether the agent is generating a report in Microsoft PowerPoint or pulling data directly into Microsoft Excel, the experience feels embedded rather than detached from real work. “Tools like Excel and PowerPoint are well-designed for solving particular problems,” says Prince. “By working within those environments, we stay accountable to the problem being solved.”
Organisations can also scale their agent strategy more reliably with LSEG’s MCP, which standardises how models discover and access data. MCP also helps uphold licensing and compliance standards, reducing governance risks while enabling rapid data access. “Originally introduced by Anthropic about 18 months ago, MCP was developed as a standard protocol so we can use the same connector across multiple places,” says Prince. “It enables models to discover underlying tools safely, while upholding standards, licensing and controls.”
For many financial services firms, this level of enablement arrives at exactly the right time. Despite years of investment in data strategy and AI experimentation, many organisations have struggled to translate pilots into scalable, repeatable value.
“A lot of customers have had the building blocks for AI but been unable to scale it,” says Prince. “So, this announcement comes at a very good moment. It’s the first ever MCP integration within Copilot Studio – a landmark that gives customers access to great, rich historical content and lets them create agents in a safe, secure way.”
The combination of democratised tooling, high-quality data and standardised access is beginning to reshape how financial institutions think about innovation.
“What used to be a linear, engineering-driven process is becoming a fluid, iterative practice that empowers domain experts directly,” says Prince. “And while agent sophistication will continue to grow, the barrier to getting started is already lower than it ever was.”
Discover more insights like this in the Winter 2025 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox.