AI Revenue Analysis: Where Should You Look When Sales Drop?

Author: AIB Team·Sunday, October 4, 2026
AI Revenue Analysis: Where Should You Look When Sales Drop?

AI sounds useful, but it can be hard to see where it fits into your work beyond a chat. Sales and purchase records, with clear figures and supporting evidence, offer a practical starting point. AI can help compare those records, highlight changes, and quickly organize the questions worth investigating next.

If revenue fell last month, for example, you can narrow down which division or customer account changed most before reviewing individual records. This article lets you try that with one revenue table, then shows how to apply the same approach using Gemini in Google Sheets or Excel analysis in ChatGPT.

1. Turn records into a table

Sales records can reveal revenue changes by division; purchase records can show changes in spending by item or supplier. Here, we will start with “Why did revenue fall?” Give AI revenue figures for two periods and ask where they changed, and by how much.

Copy the relevant accounting, sales, or operations data into three columns: revenue recognition date, group, and revenue. Choose one grouping, such as division, service, or customer account. Keep a consistent row unit, such as one day per division, with column headers and consistent date and number formats. CSV files also open in Google Sheets and Excel.

“Compare total revenue and changes by group across these two periods. Rank groups by the size of their revenue loss and show the columns and ranges used. Identify empty cells, duplicates, or other issues to check first.”

Align the periods and the treatment of cancellations and refunds, and avoid adding subtotal rows to their underlying detail. Keep contract values, invoices, and cash receipts distinct, and compare revenue recorded on the same basis. Remove customer names, contact details, and transaction identifiers from the summary you give an approved AI tool, then check the resulting totals against the original.


2. What the total hides

Now try choosing where to investigate first. Change the grouping and displayed metric, then select A, B, or C. Review the result and explore B’s operating records to build a request containing the table and your chosen record. Copy it into the AI tool you normally use to try the analysis. The exercise itself does not call AI. It uses fictional data compared on a consistent basis, with amounts in Korean won.

Find the revenue drop and build an AI question

Compare fictional revenue by division, service, or account, add an operating record, and build a request to copy into your usual AI tool.

Loading…

Total revenue fell from KRW 30 million to KRW 27.2 million, a KRW 2.8 million decline of about 9.3%. Yet A grew by KRW 3 million, while B lost KRW 4 million and C lost KRW 1.8 million. A’s growth offset KRW 3 million of the combined KRW 5.8 million loss elsewhere.

C fell from KRW 3 million to KRW 1.2 million, a 60% decline. That is a faster fall than B’s roughly 33.3%, but B lost more money. Percentage change shows the relative movement; amount change shows its monetary effect on total revenue. If revenue loss is your investigation criterion, there is a reason to start with B. Profit margins or strategic importance could change the business priority.

In your own analysis, start with one grouping, then examine the notable segment by week, product, or owner. Avoid mixing classification schemes in a way that counts the same revenue twice.


3. Questions that narrow the cause

Once you have located a large change, ask AI “What should I check next?” The table shows that B lost KRW 4 million. To narrow down why, bring in the relevant operating records.

  • Transactions and usage: Did units sold, services actually delivered, or contract renewals decline? Choose a measure that fits the work.
  • Prices and mix: Did realized prices, discounts, or the balance of higher- and lower-priced goods and services change?
  • Timing and recognition: Was delivery postponed? Did cancellations, refunds, or revenue-recognition timing change?

“Separate changes supported by this table from possible causes. Identify the records needed to test each hypothesis and choose three questions for the responsible person. Do not assert a cause that the data does not establish.”

In the exercise, a note that work worth KRW 1 million moved to the following month cannot explain the entire KRW 4 million decline. Check actual delivery dates and revenue records. An unchanged price list also does not establish that realized discounts or the sales mix stayed the same.

For sales with a well-defined order unit, order counts and average order value can help. For work measured in completed service sessions or contracts, use the appropriate definitions. Do not add unlike activity counts or force every business into an order-based formula.


4. Two practical AI options

Now take the request from the exercise into your own work. If your records are in Google Sheets, start with Gemini. If you have a CSV or Excel file, start with ChatGPT. The steps below show how to provide the data and ask for a comparison table and chart.

Gemini in Google Sheets

If your data is already in Google Sheets, start there. You need an eligible plan, such as Workspace Business Standard or Plus, or Google AI Pro or Ultra. Korean and Japanese are supported alongside English, although conditions can differ by feature.

  1. Open a copy of the spreadsheet and select Ask Gemini at the top right.
  2. Specify the sheet name or cell range, then paste the request below.
  3. Review the tables, formulas, charts, and proposed changes. Check any plan or preview shown before inserting or applying them.

“Compare revenue and changes in amount and percent by period and group. Show the segments contributing to the total decline and the growth offsetting losses elsewhere. Suggest the formulas and a bar chart of the changes.”

Gemini can help create formulas and charts and edit the sheet. Do not assume a generated chart will automatically follow changes in the original dataset. Check its source range.

Analyze and edit Excel with ChatGPT

For a CSV or Excel file, open a new ChatGPT conversation and use the + or file attachment button to upload a copy. Free accounts can start analyzing files, subject to usage limits. File editing and creation capabilities can vary by account and interface.

  1. First ask: “Read the columns and periods, and check for missing values, duplicates, and number-format problems.”
  2. Use the request below to create a comparison and chart.
  3. Open the resulting workbook in Excel and check totals against the original. Review important formulas and chart ranges too.

“Keep the original data and create a new .xlsx file with an added ‘Revenue comparison’ sheet. Summarize revenue and changes in amount and percent by period and group, and add a bar chart. Separate observed changes from possible causes. Explain the calculation rules and provide a download link.”

Start with a summary table rather than assuming complex formulas, pivot tables, or macros will be preserved. If you later need a recurring dashboard, consider Sheets canvas, currently supported in English. For connected business data, consider the Data agent in ChatGPT Work. Check the applicable plans and your organization’s connection permissions separately.


5. Choose one change this week

Finish by naming the action, its owner, the metric to watch, and a review date. If delayed delivery in B is confirmed, the owner can review the affected work and schedule. The following week, check completed delivery and recognized revenue separately. Changing fewer conditions at once makes the outcome easier to examine.

If you test a discount, track its cost and the profit left over as well as revenue. Choose an action that addresses a verified problem rather than acting on an AI-generated explanation alone.

Unsure which AI tool fits your revenue data? Start with AIB. AIB runs an independent AI comparison platform without favoring a particular AI company. It helps business owners and teams choose AI tools that fit their work and budget, and put them to use. From preparing your data to repeating the analysis, find a workflow that fits your business with AIB.

Talk to AIB about choosing and using AI


Sources and further reading