I work as a freelance spreadsheet cleaner and reporting assistant for small trading companies around Manchester, mostly wholesalers, repair shops, and family-run online sellers. I spend more hours than I admit inside Excel files that have strange column names, broken formulas, pasted bank exports, and monthly sales tabs going back 4 or 5 years. I started using ChatGPT for Excel work because I was tired of rebuilding the same formulas from memory every week. I still check the results myself, but it has changed how I plan, fix, and explain spreadsheets.
Where ChatGPT Actually Helps Me Inside Excel
The first place I use it is formula writing. A client may send me a sheet with 18 columns, no proper table names, and a note like, “Can you show late invoices by customer?” I can describe the layout in plain English and ask for a formula using XLOOKUP, FILTER, SUMIFS, or a helper column. That saves me the slow part of remembering the exact syntax, especially when the formula has 3 or 4 conditions.
I also use it for cleaning messy data before I build a report. One supplier file I handled last winter had product codes mixed with spaces, lowercase letters, and old prefixes that no one used anymore. I asked ChatGPT to suggest steps for splitting the code, trimming the spaces, and checking duplicates before I touched the real file. That gave me a safer order of work, and it kept me from fixing one problem while creating another.
It is not magic. I said that early. ChatGPT can write a formula that looks right and still miss the way a sheet is built. I have seen it suggest a clean formula for a table that was not really a table, just a pasted range with merged cells across the top.
How I Turn Plain Business Questions Into Better Spreadsheets
Most business owners I work with do not ask for formulas. They ask questions like, “Which products are slowing down?” or “Why does April look worse than March?” I use ChatGPT to translate those questions into spreadsheet tasks, then I decide what is safe to build. A simple sales file with 6 columns can often produce more useful answers than a flashy dashboard with 20 charts.
A resource like chatgpt for excel can fit naturally into that work when someone wants help turning spreadsheet questions into actual formulas or analysis steps. I still tell clients to test every answer against a small sample before trusting it across a full workbook. One shop owner I helped last spring caught a formula issue in the first 12 rows because we checked it slowly before copying it down to the whole month.
The best prompts are specific. I do not type, “make me a sales report,” because that leaves too much room for guessing. I write something closer to, “I have columns for order date, customer name, SKU, quantity, net sale, and region, and I need a monthly total by region that ignores blank SKUs.” That kind of prompt usually gives me a better starting point, even if I still change the formula after reading it.
Why I Do Not Trust the First Answer Blindly
I treat ChatGPT like a quick assistant, not a senior accountant. If it gives me a formula, I test it on 5 or 10 rows where I already know the answer. If it suggests a Power Query step, I run it on a copy of the file first. That habit has saved me from several quiet mistakes.
One recurring issue is that Excel versions vary. A formula using FILTER may work fine in Microsoft 365 but fail for someone using an older desktop version from years ago. I had a client with 3 office laptops, and only one of them could open the newer formulas properly. That was a small detail, but it changed the whole setup I gave them.
Another issue is local formatting. Dates can be awkward. I have seen day and month values flip because a file came from a supplier using a different regional format. ChatGPT can explain how to fix dates, but I still inspect the column myself and compare it with invoices or order notes before I trust the result.
How I Use It for VBA, Power Query, and Repetitive Fixes
I do not write VBA from scratch every day anymore, but I still get asked to automate dull jobs. A warehouse manager once wanted a button that copied rows marked “Ready” into a dispatch sheet, then cleared a small status field after the copy. I described the task to ChatGPT, asked for VBA, and then reviewed the code line by line. The first version was close, but it needed changes for the real sheet names and protected cells.
Power Query is another place where I find it useful. Many small companies export the same sales file every Monday, then paste it into last week’s workbook. I use ChatGPT to outline the steps for importing a folder, promoting headers, changing data types, and filtering cancelled orders. It gives me a clean checklist, which is helpful when the client’s file names are messy and the columns move around.
I do not hide the rough edges from clients. If a macro deletes data, that is my responsibility, not the chatbot’s. So I keep backups, name every test file clearly, and avoid running generated code on the only copy of a workbook. Old habits still matter.
The Prompting Habits That Give Me Better Results
I get better answers when I share the shape of the sheet instead of the whole business story. I might say there are 9,000 rows, the headers are in row 1, dates are in column A, and the customer name sits in column C. Then I explain the output I want in one clear sentence. That beats sending a vague request with no column details.
I also ask ChatGPT to explain the formula after it gives it to me. If I cannot follow the explanation, I do not use the formula yet. A few times, the explanation has revealed that the answer was solving a slightly different problem from the one I had. That is useful, because a wrong formula can look very polished.
For longer workbooks, I break the job into parts. First I ask for the cleaning plan, then the formula, then the report layout, and only after that do I ask for possible automation. This keeps the conversation focused. It also matches how I would train a junior assistant sitting next to me with a laptop and a messy workbook.
What I Tell Clients Before They Start Using It
I tell clients to use ChatGPT for Excel in places where a wrong first draft is easy to catch. Formula ideas, column cleanup steps, report outlines, and plain English explanations are usually safe starting points. Payroll, tax filings, stock valuation, and anything tied to legal reporting need a stricter review. I have no problem saying that out loud in a meeting.
I also remind people that private data needs care. A workbook with customer phone numbers, staff pay, or supplier bank details should not be pasted casually into any tool. I often replace real names with sample names and reduce a file to 8 or 10 fake rows before asking for help. The answer is usually just as useful, and the risk is lower.
Training matters more than most owners expect. I spent one rainy afternoon with a purchasing team showing them how to ask for formulas without sharing sensitive values. By the end, they were writing clearer prompts than they had written spreadsheet notes before. That small shift made their weekly stock checks less painful.
I still like Excel because it forces a person to look closely at numbers. ChatGPT makes some of that work faster, especially the parts that involve syntax, cleanup ideas, and first drafts of automation. The person using it still needs judgment, patience, and a habit of checking the answer against real rows. That is the balance I trust in my own work.
