The research tool we helped build was an important first step. To protect accuracy, we implemented an LLM-as-a-judge evaluation framework for the initial research assistant — one of the first public solutions of its kind. A secondary LLM reviewed the primary AI’s responses to help ensure that the outputs remained grounded in approved research data and within strict guardrails. It gave the client a safer way to explore AI while maintaining the trust and precision their brand depends on.

As the AI landscape matured, the firm’s enterprise clients moved quickly. They were building internal LLM-based systems and needed a direct way to integrate the firm’s research into their workflows.

That shift created a new challenge. Existing tools were still built for human interaction, requiring users to manually search, extract, and reformat information. At the same time, the firm had to navigate a major transition: moving from deterministic software into probabilistic AI while protecting accuracy, trust, and brand reputation.

Because we were already deeply embedded in the business, we were able to help the firm look beyond incremental improvements and move toward a more ambitious AI-native model.

We partnered with the firm to reimagine how its teams build and deliver AI-powered products. Instead of siloed web experiences, the organization began shifting toward a more connected agentic ecosystem.

Rather than continuing to invest only in a traditional retrieval-augmented generation, we pivoted toward a distributed network of MCP servers. This architecture exposes the firm’s specialized tools through a standardized protocol, making its data and research available to AI clients such as Claude, Microsoft Copilot, and other enterprise systems.

In practice, the firm’s products are no longer locked inside a single interface. They can travel wherever clients need them.

The solution also introduced strategic model orchestration to balance reasoning power, cost, and latency. Larger models handle more complex tasks, while smaller models support faster, more efficient requests. That creates a more scalable foundation for long-term growth.

The shift from human-centric interfaces to machine-ready data delivery has opened up a new growth channel for the firm. Early feedback has been positive, especially around the ability to embed proprietary research directly into client workflows.

Internally, the transformation has also created greater speed and flexibility. With AI-driven development tools and a more modular architecture, teams can move faster while still meeting the demands of a regulated financial services environment.

The platform is now positioned to grow with the business, with internal squads contributing specialized tools to one cohesive experience.

This was not a one-off AI implementation. It was the next chapter in a multi-year partnership built on trust, embedded collaboration, and shared ambition.

Through agentic AI and MCP, the organization can now deliver its most valuable research and data directly into the systems where clients work, build, and make decisions.

The result is a faster, more flexible, and more future-ready model for growth—one that positions the firm as a leader in AI-native financial intelligence.