You can ask questions about your own business data in plain English using tools you likely already pay for: Microsoft Copilot in Excel, Gemini in Google Sheets, or ChatGPT with a file upload. Each one turns a typed question into a formula, query, or script and returns a direct answer, without requiring SQL or a dashboard build.

Ask your business data questions in plain English, and for most Irish SME owners, the honest answer is that they have never tried because they assume it needs a data analyst or a piece of software they do not have. Most do not. Only 21.2% of Irish small enterprises use Business Intelligence software of any kind, compared with 75.5% of large enterprises, so most owners asking "what were my top five customers last quarter" are doing it by hand, or not asking at all.[1] The tools that answer that question already sit inside the software many businesses use every day.

Do you know which of your own business questions could already be answered by a tool you are paying for right now? If you are not sure what data your business actually has, in what state, an AI Readiness Scan maps that out before you go any further.

What tools let you ask questions of your business data in plain English?

They already exist inside software you very likely have open right now: the Copilot panel in Excel, the Gemini side panel in Google Sheets, and the file upload feature in ChatGPT.

All three work the same way underneath. You type a question the way you would ask a colleague. The tool converts that question into a formula, a query, or a short piece of code, runs it against your actual data, and hands back a table, a chart, or a plain written answer. None of them expect you to know a formula syntax, let alone SQL. The differences between them come down to where your data already lives and what plan or subscription you already have.

If your business runs on Microsoft 365, the answer is probably already in your Excel file. If it runs on Google Workspace, it is already in your Sheets. If neither applies, or the data is a one-off export, ChatGPT will take almost anything you hand it. The same discipline that matters for AI adoption generally, starting from a precise question rather than a tool you like the look of, applies just as much here.

In summary

Before choosing a tool, check where your data already lives. The fastest starting point is almost always the assistant built into the software you are already paying for, not a new product to evaluate.

How does Microsoft Copilot in Excel answer questions about your data?

Copilot in Excel answers a typed question about your spreadsheet directly, showing the result as a summary, a chart, a PivotTable, or a highlighted trend, and it can explain the formula it used to get there.

You will find the Copilot icon on the Home tab, or you can select a cell and use the sparkle icon that appears next to it. From there you type your question in ordinary language, something like "what's the fastest growing product line this year", and Copilot works out which columns matter, builds the calculation, and returns a direct answer alongside the working. It can also generate a new formula from a plain-English description and explain, in words, what an existing formula is actually doing, which is useful if you have inherited a spreadsheet full of other people's formulas.

The one practical requirement worth knowing before you start: Copilot in Excel only works on a file that has AutoSave turned on, which in practice means it is saved to OneDrive or SharePoint. Copilot will not work on a file sitting only on your local drive.[2]

In summary

If you already have a Microsoft 365 subscription, this costs nothing extra to try. Save your working file to OneDrive, turn on AutoSave, and ask your first question today.

How does Gemini in Google Sheets answer questions about your data?

Gemini in Google Sheets answers a typed question about your data from a side panel, and can also build the pivot table, chart, or filter that goes with the answer, all from the same plain-English prompt.

Open the sheet, click Ask Gemini in the top right, and either pick one of the suggested prompts or type your own, such as "identify trends in this table" or "help me understand month to month costs". Gemini can summarise the whole sheet, spot outliers, and generate a chart or pivot table on request, and you can check its "Analysis steps" to see how it got to its answer before you insert anything into your spreadsheet.

The catch is that this only works natively inside a Google Sheet. If your data is an Excel file, you need to open it and choose File, then Save as Google Sheets first. You will also need an eligible Google Workspace or Google AI plan; it is not switched on by default for every account.[4]

In summary

Gemini shows its analysis steps before you commit anything to your sheet. Get into the habit of checking that panel, not just the headline answer, before you insert a result.

How does ChatGPT let you ask business data questions in plain English from an uploaded spreadsheet?

ChatGPT answers a question from an uploaded spreadsheet by running it through a sandboxed Python environment, writing and executing the calculation itself, and returning the result as a table, a chart, or a written answer, with the underlying code available if you want to see it.

You upload a CSV, XLSX, or similar file directly into a chat and ask your question the way you would ask a person: "which customers had the biggest drop in orders this quarter". ChatGPT inspects the file, works out what needs to be calculated, writes the code to do it, runs it, and gives you the answer. Unlike Copilot or Gemini, it does not need Excel or Google Sheets at all. It is software-agnostic: any structured export will do.

The trade-off is scale and structure. Very large files, roughly above 50MB, or workbooks with several unrelated tables crammed onto one sheet, can push past what ChatGPT handles reliably. OpenAI's own guidance is to use clear column headers and one record per row, and to avoid mixing several tables on a single sheet.[5]

In summary

ChatGPT does not care what software your data started in. If you have a messy one-off export nobody has tidied up, cleaning the headers and splitting mixed tables first is worth more than any clever prompt.

How do you know if the answer is actually right?

You know an AI tool's answer to a data question is right by checking the working it shows you and spot-checking at least one key number by hand before you act on it, not by trusting the first answer that comes back.

All three tools show their reasoning if you ask: Copilot can explain the formula it built, Gemini has an Analysis steps panel, and ChatGPT can show the Python code it ran. OpenAI's own guidance to users is explicit on this point: review the generated code, outputs, and assumptions before relying on the result, and ask the tool to use a specific method, column, or grouping if the first attempt does not match what you actually meant.[5] The failure mode to watch for is not an obvious error. It is a confident, well-formatted answer built on a wrong assumption, most often in a calculation with more than one step, such as a percentage of a filtered subset or a total built from several smaller totals.

A five-minute habit fixes most of this. Ask the tool what it assumed. Pick one number in the answer you can verify quickly by another route, a total you already know, a customer you can check by eye, and confirm it matches. If it does not, ask the tool to show its method rather than simply asking again.

In summary

Treat the first answer as a draft, not a fact. Ask to see the working, then spot-check one number you already know before you act on the rest.

What should you check before uploading business data to any of these tools?

Before uploading any file to Copilot, Gemini, or ChatGPT, check whether it contains customer, employee, or supplier personal data, and if it does, check what that tool's data processing terms say before you upload it.

This is a data protection question, not an AI-specific one. Connecting a spreadsheet containing personal data to a cloud AI tool raises the same GDPR questions as any other cloud processing arrangement. Where a question does not actually need row-level customer detail, an aggregated or anonymised export answers it just as well and removes the exposure entirely. For the full picture of what to check in a tool's data handling terms and where EU data residency actually matters, see EU data residency and AI tools: what every Irish SME needs to know.

If asking one-off questions like this becomes something your business does every week rather than occasionally, it is worth knowing that free, ready-made workflow templates already exist for turning this into a standing tool rather than a repeated manual upload, including templates built specifically to chat with a Google Sheet or a database in plain English.[6] That is a bigger step than most businesses need on day one: once a tool starts acting on its own rather than answering a question you asked, it becomes an AI agent, with its own governance considerations. For a structured way to decide which of your business tasks are worth automating in the first place, see how to run a task audit before choosing an AI tool.

In summary

Check a tool's data processing terms before uploading personal data, and reach for an aggregated export first if the question does not need row-level detail.

Getting a straight answer from your own numbers should not depend on remembering a formula or waiting on whoever built the last spreadsheet. If you want help figuring out which of your own business questions are worth setting up this way, and which tool actually fits your data, an AI Use-Case Discovery Workshop is built around your own tools and questions rather than a generic list.

This article is for informational purposes only and does not constitute legal advice.