← Resources
Tutorial · How-to · 10 min read

Rebuilding the same report every month with ChatGPT in Excel and PowerPoint

May 31, 2026

Yes, ChatGPT can fill an Excel workbook and build a PowerPoint deck from your exports, using two add-ins installed in Office. It does not check your totals for you. This tutorial sets up the recipe in five steps for a business that rebuilds the same report every month: a small agency, a freelancer in bookkeeping or finance, an online shop.

Here is the made-up example I keep from start to finish. A four-person web agency sends six clients a monthly update (site visits, inquiries received, ad spend). For each client, someone opens three exports, copies figures into a workbook, recalculates two totals by hand, then rebuilds the same deck with the new numbers. I count one hour per client, so six hours at the start of every month. That is an illustration, not a measurement: swap in your real time before you decide anything. The other durations in this tutorial are the same kind of thing, my own estimates.

At a glance

For whom
Small agencies, freelancers in finance and online shop owners who rebuild the same report every month
You need
Excel and PowerPoint (Microsoft 365), the ChatGPT add-in and your last report
Time
Half a day to build the recipe, then less each month (my estimate)
Cost
Plus plan at around 20 dollars a month, free plan to practice on dummy figures (based on a third-party comparison read on November 1, 2026)
Level
Beginner
Updated on
November 1, 2026

On the tools side, here is what the announcements say. The Excel add-in left beta in early May 2026: it builds spreadsheets from scratch, fixes formulas and cleans up inconsistent data. The PowerPoint add-in has been in beta since May 21, 2026: you describe what you want and it creates, edits and reorganizes slides. Both are available from the free plan up to Enterprise, though actual access depends on the plan and on your Microsoft 365 admin settings.

Frame the report before touching the AI

If you open ChatGPT and type "make me a nice report", you get something generic and unusable: your report follows rules the AI cannot guess. Before automating anything, lock down the structure.

Concrete tool: a plain text document (Notion, Word, an email to yourself). No cost.

Setup (about 30 minutes the first time):

  1. Pull up the last report you sent, the real one, the one your client received.
  2. List its blocks in the exact order, with the format of each figure.
  3. Note where each piece of data comes from: which export, which tab, which column. This is the boring part, and it is what makes the rest hold.

For the made-up agency, the recipe fits in a table:

BlockFormatSource
Site visitsWhole number, change in % against the previous monthAnalytics export, "Traffic" tab, column B
Inquiries receivedWhole numberContact form export, "date" column
Ad spendEuros, no decimalsAd platform export, "cost" column
CommentaryThree sentences at mostWritten by the project lead

The same table gets filled differently elsewhere. A freelancer in finance would put the month's cash position and overdue invoices in it, an online shop owner would put sales by product and return rate.

Here is the brief template to keep in your document, which you will paste into ChatGPT in the next steps:

Prompt
Monthly report for [client name]. Blocks, in this order:
1. [block 1]: [format], source [export, tab, column]
2. [block 2]: [format], source [export, tab, column]
Summary commentary: [length, tone].

Expected outcome: a page that describes your report like a recipe. Without it, the add-in suggests a different format every time you run it, and you spend more time fixing than saving.

Line drawing: a report page with a bar chart and lines of text, a blue arrow, a sheet of six numbered blue points, a robot in between.
Your last report, described block by block, becomes the recipe you paste into ChatGPT every month.

Install ChatGPT inside Excel and PowerPoint

Both add-ins open in a panel to the right of your file. You write in plain language and they act on the open document.

Concrete tool: Microsoft 365 (Excel and PowerPoint, app or web) and the ChatGPT add-in. On the ChatGPT side, the Plus plan costs around 20 dollars a month according to this plan comparison, price taken from a third-party comparison read on November 1, 2026. Prices move, so look at the official page before you subscribe. Business is billed per user, Enterprise by quote.

Setup (about 10 minutes):

  1. In Excel, open Insert then Get Add-ins (menus vary with your version).
  2. Search for "ChatGPT", install the add-in, then connect your OpenAI account.
  3. Do the same in PowerPoint.

Expected outcome: a chat panel inside both tools, with no back-and-forth to your browser.

Fill and calculate the Excel file

Instead of copy-pasting columns and writing formulas by hand, you describe the calculation you want and the add-in lays it down. According to its announcement, it fixes formulas and cleans up inconsistent data. I have no source on how much data it handles, so test it on your own export before you rely on it.

Concrete tool: the add-in in Excel, on the raw export from your billing tool, your analytics or your shop.

Setup (about 10 minutes):

  1. Open your export in Excel and select the relevant range.
  2. In the panel, give a precise instruction, like the first prompt below.
  3. Check the result, then ask for an adjustment, like the second prompt.
Prompt
In a new sheet, calculate [the measure: visits, revenue, inquiries] by [the segment: channel, client, product] for [the current month], compare it with [the reference period] and show the gap as a percentage with one decimal.
Prompt
Sort from the largest [segment] to the smallest and turn negative gaps red.

Expected outcome: a summary sheet along the lines of this one (made-up figures):

ChannelVisits in OctoberVisits in SeptemberChange
Search4,2103,890+8.2%
Social media1,1501,290-10.9%
Direct980940+4.3%
Total6,3406,120+3.6%

The add-in explains the formulas it lays down, so someone else can read them and take over if the format changes. In the agency example, the copy-paste part, the longest of the hour, becomes proofreading.

Generate the PowerPoint from the data

Once your Excel is clean, you move to the deck. The PowerPoint add-in creates, edits and reorganizes slides from what you give it, including the summary sheet you just built. Two precautions. The add-in is in beta and OpenAI warns it can delete content by mistake, so work on a copy of the file. Complex formatting and custom fonts are not fully supported either.

Concrete tool: the add-in in PowerPoint, your summary sheet from step 3 and the recipe from step 1.

Setup (about 10 minutes):

  1. Open a copy of your usual reporting template, or a blank PowerPoint, and launch the panel.
  2. Give the brief, like the first prompt below.
  3. Then adjust slide by slide, with the other two.
Prompt
Create a monthly reporting deck in [number] slides: [the list of slides, in the order of your recipe]. Use the figures from the sheet [name of your summary sheet].
Prompt
Lighten slide [number], it is too dense: switch it to three columns.
Prompt
Rewrite the title of slide 1 so it mentions [the month] and [the year].

Expected outcome: a structured, filled deck that you proofread slide by slide. The time saved depends mostly on the deck template you provide, so I am not giving you a figure.

Connect the sources and turn it into a ritual

The announcements mention an integration with Gmail, Outlook and SharePoint. I do not rely on it here: I have not checked that it fetches an export from a shared folder by itself. In this recipe, the person running the ritual drops the export in place. What saves time is that everything else is written down.

Concrete tool: a shared folder (OneDrive, SharePoint, Drive) and a recurring reminder in the calendar.

Setup (about 15 minutes):

  1. Store each source export in the same place, under the same file name from one month to the next, for example traffic-export-client.xlsx.
  2. Put the recipe from step 1 in the same folder.
  3. Set a reminder on the first working day of the month: "Monthly report". The person opens Excel, runs the prompt from step 3, then the one from step 4, proofreads everything and sends.

Expected outcome: an identical format from one month to the next, because the recipe is locked. The day the usual person is away, anyone on the team picks up the ritual with the same recipe.

Line drawing: a hatched folder with an export sheet coming out, an arrow to a calendar whose first box is circled in blue, a clock with a small hatched sector.
The export sits in the same place, the slot is set on the first working day, the recipe says what to do next.

Final checklist

0 of 7

And now?

You just turned a start-of-month chore into a ritual someone else can run. That is a single report. The same recipe fits the weekly sales update, a client's cash tracking, the quarterly review of a portfolio. An office task that comes back every month and always follows the same rules is a good candidate.

Start with the report that eats the most hours and whose format does not change from one month to the next. A bad first pick does not cost money, it costs momentum, and momentum decides whether you automate a second one. Invoice reminders have their own rules, which I cover in automating the follow-up of your unpaid invoices.

The next step is to spot the other chores of the same kind: 12 AI automations and where to start gives you a starting list. I share this kind of how-to in my newsletter, you can sign up right below.

Get the AI news

Sign up and you get the guide "The useful AI automations in 2027", then at most one email a week with my new articles and tutorials.

At most one email a week, plus the guide for free. Unsubscribe in one click, the guide stays yours. Privacy

Related articles