.webp)
Table of Contents
Key takeaways
Claude for Excel released recently, and it works impressively well.
Unsurprisingly, that’s meant we’ve been fielding questions from people wondering why they couldn’t just use Claude or ChatGPT on top of Excel to support their planning needs.
The honest answer is that Claude inside Excel will offer a real productivity gain. Anyone using it for quick reporting, ad hoc analysis, a single-user model, or a one-off prototype should keep using it - it’s the right tool for that job.
But the moment a planning process becomes recurring, multi-user, cross-functional, or decision-critical, the job changes. Speed of building stops being the bottleneck - it becomes the burden of running the same process every month with consistent logic, trusted data, parallel scenarios, contributors across departments, and a clear audit trail when the CFO, board, or auditor asks where a number came from.
That's the work Pigment is built for, and it's also the work where LLMs become dramatically more powerful when they sit on top of a real planning platform.
What an LLM can do inside Pigment that it can't do inside a spreadsheet
When an AI is operating on a structured planning system rather than a grid of cells, it’s able to perform dramatically more complex tasks. Things like:
1. Editing live models safely
Our Modeler Agent reads and writes the model directly, in the same language a human modeler uses. It can branch the model, propose changes, run them through Test & Deploy, and roll back if necessary.
It can refactor a forecast across dozens of dimensions and prove it’s not broken anything.
2. Running scenarios across many contributors at once
Forecasting at enterprise scale means more than one user iterating in a tab. It requires pulling actuals from multiple systems, hundreds of assumptions, several parallel scenarios, and contributors across finance, sales, and operations.
Our Analyst Agent operates inside that shared, governed environment. The output is a number every contributor sees in the same place, not a file someone has to merge later.
3. Connecting to the rest of your AI stack
Through MCP, any external agent can read from and write to Pigment. That's why the Supercell setup works: Claude becomes useful across the company precisely because Pigment is the place where the planning data, logic, and access rights already live.
The LLM doesn't have to reconstruct context every time because the platform provides it.
What this looks like in practice
We give customers three ways to bring AI into their planning work:
- The Analyst and Modeler Agents, built natively on the platform, the modeling language, and planning best practices
- Custom Agents, built on the same capabilities and tailored to a specific team or workflow
- Direct access via MCP, so the AI tools your teams already use can operate on Pigment data and capabilities
Every organization picks a different mix. That's the point. The platform is designed to be the substrate, the customer decides what to build on top.
What to watch for
LLMs are a fantastic tool. But as anyone with a mature AI deployment will tell you, at company scale they need a system of record to operate on - and planning is one of the most obvious cases where that’s true.
When the planning platform is AI-native, every model improvement, scenario, and audit trail makes the next AI interaction more accurate. When the AI sits on a spreadsheet, the work resets every time someone opens up a new file.
That’s why we’re invested in making Pigment the most capable place in the company for an AI to do planning work.
To see what that looks like for your organization, join the next live tour or request a demo.
