AI for Excel, Data Analysis and Workplace Research

Two of the biggest time drains at work are wrangling data and hunting for information. In Microsoft's research, 62% of workers say they struggle with too much time spent searching for information during the workday, and the average employee spends well over half the week on coordination rather than the work they were hired to do (Source: Microsoft, 2023 Work Trend Index). AI is built to take that load off your plate: 90% of workers who use AI on the job say it saves them time (Source: Microsoft, 2024 Work Trend Index).

This is a do-it-yourself guide. By the end, you will know exactly how to use AI in three areas, spreadsheets, data analysis, and workplace research, with the specific tools, the steps to follow, and copy-ready prompts. You do not need to code, and you do not need to be a data analyst. You do need one habit that runs through this entire guide: let AI do the heavy lifting, then check its work. These tasks are exactly where AI is most useful and also where it can confidently get things wrong.

A spreadsheet formula that looks right can still be wrong, and a statistic an AI hands you can be invented. The skill is not avoiding these tools. It is using them to move faster, then verifying the parts that matter.

A Quick Word on How These Tools Think

Generative AI predicts the most likely next words rather than retrieving verified facts the way a database or search engine does (Source: IBM). For data and research work, that single fact matters twice over. The model can write a formula that produces a plausible but wrong number, and it can state a statistic or cite a source that does not exist.

That is not a reason to skip these tools. It is the reason the workflow below always ends with a check. With that in mind, let's start in the place most people live: the spreadsheet.

Part 1: AI for Excel and Google Sheets

AI inside a spreadsheet can write and explain formulas, clean messy data, build pivot tables, create charts, summarize what your numbers mean, and categorize text like customer feedback. Here is how to do each, by tool.

Method 1: Copilot in Excel, step by step

Copilot is the AI built into Microsoft 365. To use it in Excel, you need a Microsoft 365 Copilot license, AutoSave turned on with the file saved to OneDrive, and your data organized as a clean table with headers (Source: Microsoft). Then follow these steps:

1. Open the workbook and click anywhere inside your data table.

2. On the Home tab of the ribbon, click the Copilot button. A panel opens on the right.

3. Type a request in plain language and press enter.

4. Review what Copilot produces, and refine with a follow-up if needed.

Copilot in Excel offers three modes, edit, plan, and chat. Edit mode makes changes directly in your workbook, plan mode shows you a step-by-step plan to approve before it acts, and chat mode answers questions and gives insights without changing your sheet (Source: Microsoft). Try prompts like these:

• "Summarize total sales by product category for Q1."

• "Calculate total revenue for orders above $500 in the Northeast region." (Copilot writes the SUMIFS formula for you.)

• "Create a pivot table of revenue by region and month."

• "Make a bar chart of revenue by region."

• "Highlight any outliers in the Amount column."

• "Explain what this formula does."
  

A good habit is to ask Copilot to explain the formula it writes, so you understand it before relying on it.

Method 2: The =COPILOT() function, step by step

Excel now has a function that lets you call AI directly inside a cell, the same way you would write =SUM(). You type a plain-language instruction and reference your data, and the result spills into the grid. Because it is built into Excel's calculation engine, the output updates automatically when your data changes (Source: Microsoft).

To use it, click a cell and type a formula like this:

• =COPILOT("Classify this feedback as Positive, Neutral, or Negative", D4:D18)

• =COPILOT("Summarize the main complaint in each row", B2:B50)

It is ideal for categorizing text at scale, such as tagging hundreds of survey responses by sentiment, work that used to mean exporting your data to another tool. Microsoft notes that data sent through the COPILOT function is not used to train its AI models. The clearer your instruction, and the more precisely you point to the cells and the format you want, the better the result.

Method 3: Gemini in Google Sheets, step by step

If your organization runs on Google Workspace, Gemini is the equivalent, and it works best with native Google Sheets files. (If your file is an .xlsx, open it and choose File, then Save as Google Sheets.) You need an eligible Workspace or Google AI plan. Gemini lives in two places in Sheets (Source: Google Workspace):

1. The Ask Gemini side panel. Click the Ask Gemini star icon in the top-right corner. In the panel, you can ask it to summarize a range, write or explain a formula, build a chart, or surface insights. For example: "Summarize this data and tell me the top three trends," or "Write a formula to calculate the running total in column E."

2. Help me organize. On a blank sheet tab, type a description of what you want to build, such as "Weekly content calendar with date, channel, status, owner, and notes," and Gemini generates a formatted table with headers and sample rows.

Sheets also has an =AI() function that works inside a cell for row-by-row tasks, for example =AI("analyze sentiment", A2), useful for classifying or summarizing a whole column at once.

Method 4: ChatGPT or Claude for formulas, no license needed

You do not need a paid spreadsheet add-on for quick formula help. You can open ChatGPT or Claude and describe what you want, then paste the result into your sheet:

• "Write an Excel formula that counts how many cells in column B are greater than 1,000."

• "Here's a formula I'm using. Explain what it does and tell me if it could break." (Paste the formula.)

• "Write a Google Sheets formula to pull the domain out of an email address in cell A2."

One caution: a personal, free ChatGPT account is not the place for confidential company data, a point we return to at the end. Keep your inputs generic, or use your organization's approved account.

Getting good results in spreadsheets

A few habits make every method above work better. Format your data as a clean table with clear headers before you start, since merged cells, inconsistent dates, and blank columns confuse the AI. Be specific about the columns, the calculation, and the output you want. And always check the result: spot-check a formula on a row where you know the answer, and make sure a chart actually reflects your data before you share it.

Part 2: Data Analysis with AI

Beyond individual formulas, AI can analyze an entire dataset and tell you what it means, often in seconds. Analyzing and summarizing information is among the most common business uses of generative AI (Source: McKinsey, The State of AI). This is where the time savings get dramatic, and where verification matters most.

How AI Is Transforming Data Analytics

The standout tool: ChatGPT Advanced Data Analysis, step by step

ChatGPT can take a data file, analyze it, and build charts without you writing a single line of code, a feature often called Advanced Data Analysis (Source: OpenAI). Here is the full process:

1. Export your data from your system as a CSV or Excel file.

2. Open ChatGPT and click the attachment or upload icon (the paperclip or plus near the message box). Select your file.

3. Ask a question in plain language about the data.

4. Refine and dig deeper with follow-up questions, and download any chart it produces.

These prompts will get you started:

• "Summarize this dataset in five bullet points, including the size and the main columns."

• "Which three regions grew fastest last quarter? Show me a bar chart of revenue by region."

• "Are there any outliers or unusual patterns I should know about?"

• "Plot monthly revenue over the past year and describe the trend in plain language."

• "Is there a relationship between marketing spend and sales in this data?"

A useful move is to add "show me your steps" so you can see how it reached a number, which makes errors easier to catch.

A worked example, start to finish

Say you have a CSV of last year's sales, with columns for date, region, product, and amount. Here is a realistic five-step session you can copy:

1. Upload the file and ask: "Summarize this dataset and tell me what each column contains."

2. Then: "Show total revenue by region as a bar chart, sorted highest to lowest."

3. Then: "Which month had the highest and lowest revenue, and by how much?"

4. Then: "Flag any regions or months that look unusual compared to the rest."

5. Finally: "Write a three-paragraph summary of these findings for my manager, in plain business language."

In a few minutes, you have gone from a raw export to charts and a written summary. Inside a spreadsheet, you can do much of the same: Copilot in Excel and Gemini in Sheets both read the data you already have and answer plain-language questions about it, surfacing trends, outliers, and the story behind your totals.

The most important caution: AI can be confidently wrong with numbers

Here is the finding to keep in front of you any time you use AI for analysis. In a field experiment with 758 consultants, AI improved their work on suitable tasks, but on a task that required careful analysis of facts and figures to reach a correct answer, the group using AI actually performed worse than the group without it (Source: Harvard Business School). They trusted a confident, wrong result.

Data analysis is the single area where AI is most likely to mislead you while sounding completely sure of itself. Treat every number as a draft. Recompute a key figure yourself, spot-check it against the raw data, and ask the tool to show its method before you act on the result.

So build a verification step into your routine. Pick one important number the AI gives you and check it by hand or with a formula you trust. If the analysis will inform a real decision, that check is not optional.

Part 3: Workplace Research with AI

Finding and making sense of information is one of the most common reasons workers turn to AI chatbots in the first place (Source: Pew Research Center). Done well, AI can compress hours of reading into minutes. Done carelessly, it can hand you confident misinformation. The tool you choose makes the biggest difference here.

Perplexity, step by step

For research where facts matter, a tool that searches the live web and shows its sources is the safer starting point. Perplexity is built for exactly this:

1. Go to Perplexity and type your question.

2. Read the answer, which comes with numbered source links.

3. Click the sources to confirm the claims before you use them.

Try: "What are the main differences between SOC 2 Type I and Type II reports? Cite your sources." You get a clear answer and the links to verify it, which is the part that protects you.

ChatGPT Search and Deep Research, step by step

ChatGPT has two research modes. Its search feature browses the live web for a quick, current answer with links. For something larger, Deep Research runs a multi-step investigation that reads many sources and produces a structured, cited report, a job OpenAI says can take from a few minutes to half an hour and do work that would otherwise take a person hours (Source: OpenAI).

To run a deep research task, turn on the Deep Research option, then give it a detailed brief: "Research the main competitors in the project-management software market for small businesses. For each, summarize pricing tiers, target customer, and standout features, and cite your sources." It will work for several minutes and return a report you can check against its citations.

Gemini for research across your own files

If you work in Google Workspace, Gemini can research across both the web and your own documents, emails, and Drive files, which is useful when the answer lives partly in your company's material. You can ask it to synthesize information from selected files and the web in one step.

A research workflow you can reuse

Whatever tool you pick, the same five steps keep your research reliable:

1. Write a specific question, including the scope and what you need (pricing, dates, features, sources).

2. Use a tool that cites sources, such as Perplexity or ChatGPT's research modes.

3. Click through to the sources and confirm the key claims yourself.

4. Synthesize the findings in your own words.

5. Verify any number, quote, or fact you plan to rely on against the original.

The verification rule

This bears repeating because the risk is real. AI can invent statistics, misattribute quotes, and even fabricate sources that look genuine. The consequences reach the real world: a company was held liable after its chatbot gave a customer false information, with a tribunal ruling that "the AI said it" is no defense (Source: American Bar Association). Never pass along a fact from AI research without confirming it at the source.

Putting It Together Safely

Across spreadsheets, analysis, and research, the workflow is the same: let AI handle the heavy lifting, then verify the part that matters. Two rules keep that workflow safe.

Protect company data. Whatever you type or upload into a public tool may be stored outside your control. Samsung restricted employee use of ChatGPT in 2023 after staff pasted sensitive internal code into it (Source: TechCrunch). For real company data, use your organization's approved business or enterprise account, which is built to keep that data protected, rather than a personal free login. The built-in tools, Copilot in Excel and Gemini in Sheets, keep your data within your organization's environment, which is part of why they are well suited to work.

Follow your company's policy and a sound framework. If your organization is still shaping how it uses AI, the voluntary AI Risk Management Framework from the US National Institute of Standards and Technology gives any size of team a structured starting point (Source: NIST). And remember that the underlying skill, working well with AI on data and information, is among the fastest-growing employers want (Source: World Economic Forum, Future of Jobs Report 2025).

You now have a working process for the three tasks that eat the most time: AI for Excel and Sheets, AI for data analysis, and AI for workplace research. Pick one, open the tool that fits it, follow the steps, and use the example prompts as a starting point. Then do the one thing that separates good use from risky use, which is to check the result, especially the numbers and the facts.

Done this way, these tools turn hours of formula-writing, data-crunching, and reading into minutes, and free you to spend your judgment where it counts. The features will keep changing and improving. The process stays the same: let AI do the heavy lifting, and keep a human in charge of the answer.

Key Takeaways

• In Excel, use Copilot's side panel for formulas, pivot tables, charts, and insights, or the =COPILOT() function to run AI inside a cell; in Google Sheets, use the Ask Gemini panel and the =AI() function.

• No Microsoft 365 Copilot license? Paste formula requests into ChatGPT or Claude and drop the result into your sheet.

• For data analysis, upload a CSV or Excel file to ChatGPT's Advanced Data Analysis and ask plain-language questions to get summaries and charts with no code.

• Data analysis is where AI is most likely to be confidently wrong, so recompute key numbers and spot-check against the raw data every time.

• For research, use a tool that cites sources, such as Perplexity or ChatGPT's search and Deep Research, then click through and confirm before you rely on anything.

• Use your company's approved account for real data, keep confidential information out of personal tools, and follow your organization's AI policy.