AI for Working With Excel: Analyzing Spreadsheets Locally, No Cloud

Spreadsheets are where half of work's routine lives. Merging two exports, finding discrepancies, computing totals by condition, pulling together a summary from a dozen files. AI helps with this, but spreadsheets have their own quirks worth knowing upfront — both what's possible and what isn't. Let's go through it.

What AI can genuinely do with spreadsheets

Starting with the useful part. Describe a task in words, and an agent can do things with spreadsheets that would otherwise eat up your time by hand:

• merge data from several files into one — pulling monthly exports into a combined summary, for instance;
• find discrepancies between two spreadsheets — where a sum doesn't match, what's missing, what's duplicated;
• compute by condition — totals, averages, filtered selections;
• draft a report based on the data and explain where the numbers came from.

The point isn't that AI replaces Excel — it's that it removes the mechanical part: instead of manually cross-checking rows and writing formulas, the agent does a first pass, and you check the result.

Why local matters especially for spreadsheets

Spreadsheets are often the most sensitive thing in a job. Payroll, customers, finances, personal data. And those are exactly the files you'd least want to upload to an online service, because the moment you upload, you lose control over them.

Doka works with spreadsheets locally: files get parsed on your own computer, and if you connect a local model, the AI itself never reaches the network either. For work data, that's not overcaution, it's a normal requirement — you get the convenience of AI without risking the spreadsheet showing up somewhere it shouldn't.

How this works in practice

The mechanics are simple from the user's side. You show the agent a working folder with files, or attach a spreadsheet, and describe the task in words: "compare these two exports and show me where the totals diverge." From there, the agent reads the files and does the work, and Doka shows the steps it took — you can see which files it touched and what it did with them.

For heavier spreadsheet operations, the agent has another card to play: the terminal. When a task needs real computation, it can process the data with a script right on your machine, instead of eyeballing it. Still entirely local.

Being honest about the limits

The picture wouldn't be complete without this. A couple of things worth keeping in mind.

First, the cleaner and clearer a spreadsheet's structure, the better the result. A tidy table with clear headers parses well; a chaotic sheet with merged cells, a header spanning half the screen, and mixed-in data — noticeably worse. That's true of any tool, not just AI.

Second, the result needs checking, especially the numbers. A model can miscount or misread a condition. The right mode is: the agent does the heavy draft work, and you keep the final check on totals for yourself. In finance and reporting, that's not nitpicking, it's a required rule.

Third, a model might not take in a very large spreadsheet in one pass — in that case it helps to ask specific questions in parts, rather than "analyze all of it at once."

Where to start

Take a routine task you do regularly — merging two exports, or finding discrepancies — and run it on real files. That's the fastest way to see where AI saves time and where you need to keep control. Download Doka for free, and for working with other formats, see the article on AI for Excel, PDF, and Word files. If your recurring task isn't a one-off merge but a work report built from exports like these every month, see the article on AI for reports built from your own files.