Urgent call for CFOs: Connect AI to everything (MCP explained)
AI grew hands, feet, and access to your ledger, finally.
Just 6 months ago, AI could talk about your business.
But today, it can touch it.
MCP (Model Context Protocol) is an open protocol developed by Anthropic to give AI agents a simple way to plug into tools, data, and services with no hacks.
Think of it like a USB-C port for AI.
It can query your ERP.
It can move your cash.
Draft your journal entries.
And soon, it will close your books on its own.
You may not have heard the MCP yet.
But every major player like Microsoft, Genesis Global, Adyen, and Modern Treasury is already deploying it. And if you work in finance or lead a team, it will hit your world faster than any trend since the cloud.
This is the moment you were waiting for.
Finally AI stopped being a copilot and became an operator.
I’m excited to explain to you in today’s letter what MCP really is and how it changes everything for CFOs.
Let’s dive in.
What is MCP, really? (And How it Works)
Before, every use case needed manual integrations.
Brittle APIs, one-off connectors, and endless IT tickets. Your AI might see the data, but it couldn’t touch it. MCP is what breaks the glass.
MCP gives AI a common language to talk to any system, safely and instantly. It’s an open protocol. A standard that lets AI connect to tools like SAP, Oracle, NetSuite, Salesforce, or your data warehouse without custom plumbing each time.
How It Works
Think of MCP as the financial nervous system for AI.
Here’s the setup:
The MCP Host is your AI. Say Claude, ChatGPT, or your internal agent.
The MCP Servers are connectors. One for your ERP, one for your CRM, and one for your treasury system.
The Protocol defines the rules. How AI requests data, takes actions, and gets results.
When the AI wants to do something… “pull last quarter’s gross margin by region,” for example. It sends that request through the protocol.
The ERP’s MCP server interprets it, fetches the right data, formats it, and sends it back in real time. Without any human middleman.
Now extend that across systems:
AI can cross-reference sales forecasts from CRM with cost data from ERP and balance sheet projections from your FP&A model.
All live, connected, and compliant. Boom!
With MCP AI can actually do things.
MCP and APIs Are Not The Same
APIs were made for developers. MCP is built for AI operators.
With MCP, finance doesn’t wait on integrations. It just works.
APIs are old-school built so one app can talk to another if you hard-code every step.
MCP is different.
It’s like plugging your AI into a universal socket that instantly connects to everything. Your ERP, email, calendar, and CRM, without extra wiring.
The best part is it’s two-way and real-time. Your AI can ask for data (like “check my schedule”) and then do something with it (like “move that meeting and email the team”)
All in one flow. It’s alive, interactive, and makes AI actually useful.
The Quiet Finance Revolution is Happening
Finance is a domain of actions, not just insights.
Emails need sending.
Budgets need updating.
Reports need compiling.
Forecasts need reconciling.
Every finance process, from consolidation to audit prep, has hundreds of small, repetitive steps. Steps that LLMs could understand but not execute.
Until now.
MCP servers act as gateways between the model and the real world.
They let the AI fetch live data, process it, and send instructions to other systems, all with security and context.
So instead of asking:
“What was our Q3 EBITDA?”
You can say:
“Pull Q3 EBITDA from the ERP, compare it with last year, and draft a variance explanation.”
And your AI will:
Connect to your ERP (via an MCP server)
Retrieve the numbers
Analyze the variance
Generate commentary
Store the result in your board deck folder
That’s no longer a chatbot.
That’s a mini CFOP with execution power.
Anthropic built the first open MCP implementation and SDK.
Read this amazing piece to understand better where we’re heading.
Claude app already integrates with MCP servers for Slack, Google Drive, Notion, and Postgres. Finance teams can connect data warehouses and chat directly with live financial data, with no retraining or manual uploads.
This Is Exciting For Finance, so What’s Next?
Microsoft quietly rolled out Model Context Protocol (MCP) across Dynamics 365 Finance & Operations. And it’s not just another update.
You must know what MCP is and how it works so you can explain it back to your CEO and colleagues. This is really the revolution and I’m excited that you’re reading this and we’re in it together.
I’m working on a couple of examples of MCP servers doing real finance tasks like booking journal entries and retrieving reports from QBO.
We’ll build agentic workflows.
I’ll share them coming Thursday but it’s only for the paid users.
You must be thinking, why pay? Well, so far we’ve been very hard on sponsors because I don’t want to spoil the quality of our letters and the reading experience we’re maintaining.
25,000+ CFOs and finance leaders from the world’s biggest companies are reading this and I want to keep finance ahead of all the noise and give a clear picture of what’s happening and where the most ROI is.
Upgrade to paid and enjoy this great piece. I wrote last Thursday.
The AI CFO Office is the community inspiring the next generation of finance teams.
Automating tasks is really now a thing of the past.
The Sunday newsletters are going to be free forever.
And that’s all for today.
See you on Sunday!
I’m Wouter Born. A CFOTech investor, advisor, and founder of finstory.ai
Find me on LinkedIn