Back to Blog
AI MCP Finance EPM

Why I'm Building MCP Servers for Finance

Edward Roske

Edward Roske

Large language models are impressive. They can reason, summarize, generate code, and hold a conversation that would have seemed like science fiction five years ago. But ask Claude or ChatGPT to pull your Q3 actuals from Oracle EPM and you’ll get a confident hallucination.

That’s the gap.

The Problem

Enterprise financial systems (Essbase, Oracle Planning, HFM, consolidation tools) hold the most critical data in any organization. Revenue, expenses, forecasts, intercompany eliminations, close status. This is the data CFOs bet their careers on.

But these systems are walled gardens. They speak MDX, REST APIs with baroque authentication, and proprietary grid formats. LLMs speak natural language. Without a bridge, you get one of two outcomes:

  1. The AI guesses, and you get plausible-sounding nonsense about your financials
  2. The AI admits it can’t help, which is honest but useless

Neither is acceptable for a finance team that needs answers at the speed of a conversation.

Enter MCP

The Model Context Protocol (MCP) is that bridge. It’s an open standard that lets AI models connect to external data sources through structured tool interfaces. Think of it as giving the AI a set of approved functions it can call: “get the trial balance,” “run this MDX query,” “check the close status,” all backed by real data from real systems.

I’m building MCP servers for Oracle EPM because I spent 25 years implementing these systems. I know the data models. I know the APIs. And I know what finance teams actually need when they sit down with an AI assistant.

What This Looks Like in Practice

Imagine asking your AI assistant:

“Show me the budget variance for revenue across all entities for September.”

Instead of fabricating numbers, the AI calls your MCP server, which queries Oracle EPM, and returns actual data from your actual system. No hallucinations. No export-to-Excel-then-upload dance. Just a question and an answer grounded in truth.

That’s what I’m building.

Why Me

I founded interRel Consulting at 22 and spent 25 years as the longest-standing Oracle Platinum EPM Partner. I’ve implemented Essbase and Planning for hundreds of organizations worldwide. I have a Master of Data Science from SMU.

I’m not building MCP servers because it’s trendy. I’m building them because I’ve watched finance teams struggle with the gap between what AI promises and what their systems can deliver. I know both sides of that gap.

What’s Next

The Oracle EPM Consolidations MCP server is launching now. Essbase and Planning servers are in development. The goal is a complete suite that lets any LLM (Claude, ChatGPT, Gemini) talk to your financial systems as fluently as it talks to you.

If you’re a CFO wondering how AI fits into your finance stack, this is how. Not by replacing your systems. By making them conversational.