Giving AI the Right Context to Understand Your Investment Process
TLDR: Investment Management is changing faster than ever before due to the proliferation of AI and its impact on investor workflows. But not all investors are using AI to the fullest. Model Context Protocol (MCP) gives investment teams a new way to use the AI tools they already rely on, without losing the context, permissions, and analytics that their organizations require.
There is no doubt that AI has changed how investment teams operate. A PM can now ask for a portfolio summary in plain English and (usually) get a decent response. An analyst can ask for an earnings preview from a chatbot that is connected to financial market data. Those types of questions usually begin from scratch, with the model not knowing anything about you, your fund, or your data.
Foundation models are getting extremely powerful but Claude, ChatGPT, Gemini, and other AI tools are trained to produce output that is fitting for the general public, like retail investors researching a stock for the first time. But when the question is about your fund, the absence of context makes these powerful models at best interesting at worst, next to useless. For it to produce usable output, the AI must be connected to the right data, have the right definitions that are sometimes unique to your fund, have the right permissions in place, and the right investment context before it answers. Easier said than done.
Thankfully for our clients, EDS is the system where all the right data lives. Given that, we are proud to introduce our MCP layer for Fusion AI.
MCP, or Model Context Protocol, gives AI tools a governed way to call EDS directly. It’s like a direct pipe from the tool you use to the data you trust. After setting it up, an investor can ask a question in the AI tool they already use and enable that tool to retrieve structured, verified information from EDS. The answer has fund-specific context, source attribution, and calculations produced by EDS analytics rather than generated by the model.
For investors this means the difference between an answer that is just "interesting" and something they can trust to make decisions. Traditional AI chatbots are very good at summarizing public information like reading a transcript or explaining a filing. But a fundamental investment team does not make decisions from public data alone. They use internal research notes, expert call notes, their own risk data, attribution, price targets, various internal models, factor exposures, broker research, and so on.
Without that context, AI gives answers that may sound plausible but are not reliable for an investment process inside an asset manager and/or multi-billion dollar hedge fund. With the EDS MCP layer, the model does not have to guess, it can retrieve the correct data, and return an answer grounded in the fund’s actual system of record.
The common brain beneath every AI interface
Our view is that AI will create more interfaces for investors to interact with their data. Some users will work inside Fusion AI and others will ask questions from Claude or ChatGPT. But the touchpoints will increase and will become more conversational.
Our clients are in a unique position to benefit from this change. EDS already brings together research, portfolio construction, risk management, performance attribution, and external data in one investment process management platform. The MCP layer extends that foundation into the AI tools investors are adopting. It lets EDS act as the semantic layer for AI native fundamental investing: the place where internal quantitative and qualitative data is organized, permissioned, contextualized, and made available for governed AI use.
For example, a PM could ask, “Give me the YTD factor and idiosyncratic return for the portfolio.” Without EDS, even a frontier model has no way of getting back with an actionable answer. Trust is eroded when an AI model returns something incorrect or irrelevant. With EDS MCP, the AI can call the appropriate EDS tool, retrieve the factor performance history, and return the exact date stamped answer that is easy to audit and verify.
Deterministic, auditable, human-led
This architecture operates within the boundaries institutional investors need. While generic LLMs synthesize and explain, EDS calculates. Every number comes from verified analytics instead of being “generated” on the fly. The results remain entitlement aware, so users only see data within their approved scope.
With Fusion AI and the EDS MCP layer, AI meets investment teams where they already work while keeping the trust, structure, and auditability of the EDS platform. Claude, ChatGPT, and other AI tools become more useful because they can now ask EDS for the answer. EDS remains the common brain that makes the answer something investment teams can trust and act on.
See Fusion AI on your data.
