You will be able to write a request that returns a correct formula for your actual sheet.
Priya writes a weekly Monday report, the same one they scored in lesson 1.4. Every week they filter the orders sheet by region, then by month, then by status, copy the totals into the report, and repeat for each region, which takes forty minutes before they write anything (Priya's figures here are example figures).
Priya knows a formula could do this, but they never remember which formula or the order of its parts. An AI assistant does this job well, because a formula is easy to check: it either gives the right number or it does not. Your part is describing the sheet well and testing what comes back.
The assistant cannot see your spreadsheet. Most bad formulas from assistants come from bad descriptions of the sheet. With a poor description, the assistant guesses where the data is, picks plausible column letters, and produces a formula that is correct for a sheet you do not have.
Start with the app. Say whether you use Excel or Google Sheets, because a few functions and behaviours differ between them. Then name the sheet and the columns, giving both the column letter and the header name for each. Say which row the data starts on and roughly how many rows there are.
Next, paste three to five sample rows, with confidential details replaced first. Lesson 2.4 shows how to do that. Sample rows let the assistant see whether your dates are real dates and your amounts are numbers.
Priya's description said they use Google Sheets and the sheet is called Orders. Headers are in row 1, data starts in row 2, and there are about 3,000 rows. Column A is Order date, column B is Region, column C is Status and column D is Amount in S$. They pasted four sample rows with made-up customer names.
Describe the result in plain language, the way you would to a colleague who is good with spreadsheets. Be exact about the conditions. A request like "total sales by region" leaves the assistant to guess which column to add up, which statuses count and which dates you mean. An exact request names the column to add up, each condition, and the date range: for example, the total of column D for one region, where Status is Paid, for orders dated in March 2026.
Say where the answer should go, and whether any condition should come from another cell. If the region is taken from a cell instead of being typed into the formula, you can copy one formula down a column and get a total for every region.
Priya asked for a formula to sit next to each region name in their report, with the region taken from that cell, so one formula copied down would cover every region. The assistant gave them a SUMIFS formula with four conditions: region, status, start date and end date.
You do not need to memorise functions. Recognising which family your request belongs to is enough to make sense of the answer the assistant gives you.
A conditional total, adding up values that meet set conditions, usually uses SUMIFS. Counts under conditions use COUNTIFS or a related function. A lookup, finding a value in another table such as a customer's account manager from a list, uses a function such as XLOOKUP. Tidying dates and text, such as splitting names or converting text into real dates, has its own set of functions.
Always ask for a one-line explanation of each part of the formula. A prompt such as "Explain each argument in one line" is enough. The explanation is how you check the logic against your actual columns and data. If it says one argument checks column C for "Paid" but your status column is B, you have found an error.
Then test the formula against an answer you already know. Work one out by hand: filter the sheet the old way for one region and one month, and note the total. If the formula gives the same number, test one more case with different conditions. If it does not match, compare the explanation with your data to find the mismatch. Lesson 5.2 covers tracking down why a formula gives the wrong answer.
Priya tested the formula on North for March against the total they had filtered by hand the week before, and it matched. They then tested a region that had a cancelled order that month, to confirm the Status condition was working, and it matched again.
Most everyday formulas, including SUMIFS, COUNTIFS and IF, work the same in both apps. XLOOKUP exists in current Excel and in Google Sheets, but not in older desktop versions of Excel, where INDEX with MATCH does the same job. Google Sheets also has functions Excel lacks, such as QUERY.
Both apps follow regional settings, so the same typed date can be read differently depending on the file. 03/04/2026 may mean 3 April in one file and 4 March in another. Some regions also use semicolons instead of commas between function arguments. Telling the assistant which app you use and pasting real sample rows avoids most of these problems.
Pick three calculations you still do by hand. For each one, write the layout and the result in words before you open the assistant.
Describe three calculations you do manually today, get a formula for each, and test it against an answer you already know.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).