Using LLMs to Review and Understand Excel Models

One of the most effective uses of large language models (LLMs) is helping people get up to speed on unfamiliar code.

Excel models don’t get that benefit.

As spreadsheets grow, logic becomes harder to reason about:

• formulas are scattered across sheets,
• cell references hide intent,
• charts obscure dependencies,
• and sharing .xlsm files is often risky or impractical.

If you didn’t build the model — or if enough time has passed — understanding how it works can be surprisingly slow.

---

A simple shift in approach

Instead of asking an LLM to build or modify a spreadsheet, I focused on something simpler and more reliable:

Export the Excel model into a clean, text-based representation that both humans and LLMs can review.

The exporter produces a deterministic, read-only artifact that captures:

• sheet structure,
• formulas,
• chart bindings,
• and VBA (if permitted),

all as plain text.

Nothing executes.
Nothing is modified.

At that point, an Excel model starts to look a lot more like a codebase — and LLMs become useful reviewers instead of guessers.

---

Why this works

LLMs aren’t good at clicking through spreadsheets or interpreting screenshots.

They are good at reading structured text, understanding relationships, and summarizing behavior.

By converting Excel into the kind of artifact LLMs already handle well, you stop fighting the tool and start using it for what it’s good at.

This also helps humans: reviewing structure and formulas side-by-side in text form is often clearer than navigating a live workbook.

---

A practical workflow

In practice, this enables a cleaner loop:

1. Build or modify an Excel model
2. Run the exporter
3. Review the exported artifact (yourself, with collaborators, or with an LLM)
4. Return to Excel only after structure and logic are clear

---

Code and demo

• GitHub: exporter code and a public example
• YouTube: a short demo showing an LLM reviewing an exported model

---

Closing thought

LLMs aren’t a shortcut to understanding.

But when you give them the right artifacts, they can dramatically reduce the time it takes to get oriented.