An AI Data Analyst for Your Spreadsheets

"Why did gross margin drop in March?" is a question a spreadsheet can answer and most people can't, at least not quickly. An AI data analyst in Damon takes the file, works out what's in it, answers in words with the numbers that support the answer, and gives you a chart you can put in a dashboard. Then, because the same question comes back every month, it reruns on a schedule.

Step 1: Put the data where the agent can read it

The Damon drive is the agent's native file system. Drop the spreadsheet (CSV or Excel export) into a folder, say Finance/monthly-pnl.csv. If the data lives in Google Drive, connect it and the agent can read it there instead, but the Damon drive is the simplest starting point and needs no integration.

Step 2: Ask

In chat (the default agent is fine; a dedicated @analyst with "show your work" in its personality is better):

Using Finance/monthly-pnl.csv, why did gross margin drop in March? Compare March with February by product line and by cost category, and tell me which lines explain most of the change. Chart margin by month for the year.

You get a written answer, the table that supports it, and a line chart. The answer cites the file, so the numbers are traceable. Ask follow-ups in the same thread: "Exclude the one-off inventory write-down and recompute."

Step 3: Keep the chart

Charts the agent makes are views: saved visualizations bound to a data stream. Ask it to add the margin chart to a dashboard called "Finance" and it becomes a panel you can revisit, alongside stat cards (this month's margin), gauges (margin versus target) and tables. Views support transforms such as rolling averages and simple forecasts, so "add a 3-month rolling average" is a one-line request.

Step 4: Make it recurring

Ask: "Every month on the 3rd, rerun this on the latest file in Finance/, update the dashboard, and message me a three-line summary with anything that moved more than 2 points." Damon builds a scheduled workflow — trigger, agent step, message step — and the analysis becomes a monthly report you read rather than produce.

Save the analysis conventions as a skill (which columns are revenue, how you define margin, what "material" means) so every finance question the agent answers uses the same definitions.

What it's good at, and what it isn't

Good at: aggregation, comparison, variance explanations, cohort splits, finding outliers, reshaping data, charting, and doing all of that again next month without being asked. It's also useful for the boring part: "which rows have a missing cost category?"

Not a replacement for: statistical modeling you'd defend in a board meeting, or judgment about what the numbers mean for the business. It will tell you margin fell because product line C's COGS rose 18%; deciding what to do is yours.

Variants

• Database instead of a file. For data that lives in a warehouse, ask us about connecting it; the questions stay the same.
• Sales analysis for e-commerce. Weekly export from your store, questions about conversion by channel, chart by week.
• Campaign reporting for marketing. Ad platform exports in a folder; the agent reconciles spend and results across them.

Common questions

What file formats work?

CSV and Excel exports in the Damon drive, and spreadsheets in a connected Google Drive.

Does it modify my spreadsheet?

No. It reads the file and writes its outputs (tables, charts, summaries) as new files and views. Any write to the drive is a file.write action, gated by default.

How big a file can it handle?

Spreadsheet-sized files are comfortable in a chat request. For very large tables, a SQL data source is the better route.

Which model does the analysis?

Your choice of Anthropic, OpenAI or Google models, through zero-data-retention endpoints. The data isn't kept by the provider.