Skip to main content
USE CASE

From Excel to AI. How to get value from the master spreadsheet you cannot rip out

Every operator over 30 has this Excel. Every consultant wants to throw it away. That is wrong. The Excel is a 15-year working memory of how the business actually runs.

Your Excel is not the problem. The problem is that nobody wrote down what the tabs mean and the formulas only one person understands. I map the structure, define the columns and tabs in plain language, then build AI on top of it. The Excel keeps running. The AI reads it, queries it, and produces the outputs the team used to hand-build.

"Esta es nuestra Biblia aca"

At an industrial parts distributor in the US, the operations lead pointed at a single Excel pivot table powering a multi-million-dollar book of business and said "esta es nuestra Biblia aca." At Concreto in Argentina, an engineer described his cost workbook as "el Excel ese que te explota la vista cuando la abris."

At another construction firm, the team admitted "tambien se confunden, ustedes mismos, se confunden en el Excel." A logistics operator told me her billing run is "un dia entero" per customer because the inputs are scattered Excels.

Every operator over 30 has this Excel. Every consultant who shows up to "modernize" wants to throw it away. That is wrong.

WHY NOT REPLACE

Why ripping it out is wrong

The Excel is a 15-year working memory of how the business actually runs. The formulas encode logic nobody documented. The tabs encode workflow. Ripping it out means losing institutional knowledge and triggering an internal revolt. It also wastes six months on something that does not need to change.

What needs to change is the layer on top.

STEP ONE

Mapping

I sit with the person who owns the Excel and we name every tab in plain language. What is "cascada"? What does the "anotaciones de Armando" tab actually contain? Half the time the owner laughs because nobody ever asked. We write a one-page legend.

STEP TWO

Source of truth

If a number lives in three tabs and disagrees, we pick one. If a calculation depends on a manual paste from another file, we either automate the paste or accept the pattern and document it. This step takes a week. It produces no AI yet. It is the foundation.

STEP THREE

The Skill

I build a Claude Skill that knows the structure of the Excel. It can read it, query it, generate dashboards, draft reports, answer questions in plain language. The Excel does not change shape. The owner keeps editing it the way they always have. The Skill rides on top. The Excel stays yours, the Skill stays yours.

STEP FOUR

The second Excel

Most clients have more than one Excel. A costs workbook, a contracts log, a quotes pipeline. Once the first Skill works, I connect the second. Then the third. Within a quarter the team has a unified view across files that used to require ten minutes of clicking to assemble. If the Excel feeds a recurring report, this is where automated reporting picks up. If data also arrives by message, the email and WhatsApp ingest pairs with this.

I have done this for a $50 million distributor, a 45-fund LATAM real estate manager, a packaging manufacturer with one ERP per country, and a small construction firm. The Excel does not die. It gets useful. See the Excel-master pattern in construction and the Excel-Bible pattern in distribution. South Florida operators can read AI for construction companies in Miami for the local cut.

If the team trusts the Excel, the AI inherits that trust by reading from it.

Excel Questions

Frequently Asked Questions

Your Excel is not the problem. Nobody wrote down what the tabs mean and the formulas only one person understands. I map the structure, define the columns and tabs in plain language, then build AI on top of it. The Excel keeps running. The AI reads it, queries it, and produces the outputs the team used to hand-build.

Yes. I sit with the person who owns the Excel and name every tab in plain language. We pick one source of truth where numbers disagree. Then I build a Claude Skill that knows the structure. The Excel does not change shape. The owner keeps editing it the way they always have. The Skill rides on top.

You do not break it. The Excel is a 15-year working memory of how the business actually runs. Ripping it out means losing institutional knowledge. The mapping phase produces a one-page legend. The Skill phase adds the AI layer. Nothing in the Excel changes. The team trusts what they already trust.

Four steps. Mapping (one week, name every tab). Source of truth (one week, no AI yet, foundation work). Skill build (Claude reads the Excel, queries it, drafts reports). Second Excel connected (costs, contracts, quotes pipeline). Within a quarter the team has a unified view across files.

Most clients have more than one Excel. A costs workbook, a contracts log, a quotes pipeline. Once the first Skill works, I connect the second. Then the third. Within a quarter, queries that used to require ten minutes of clicking across files run in seconds because the Skill knows all three.

Want to see what the legend and the first Skill look like before you commit? Book a 30-minute call. I will show you an anonymized one from a current client.

You will see the tab map, the source-of-truth decisions, and the Claude Skill that reads the Excel.