Weekly KPI report in ChatGPT: paste this week's numbers, download the PowerPoint

If you build the same KPI deck every week — sales attainment, operations metrics, a finance review — the SlideForge app in ChatGPT turns this week's numbers into a native, editable .pptx on your own company template in about a minute. Start the message with @SlideForge, paste the numbers, and describe the slides. Your figures are bound verbatim, the layouts come from a catalog of scoreboards and charts built for exactly this deck, and next week you ask for the same deck again with new numbers.

The Monday deck

A weekly report is six to eight slides that never change shape: headline numbers against target, a trend, a breakdown by region or team, a status board, what happens next. What changes is the numbers, and the numbers are where copy-paste goes wrong. Most weekly report decks on SlideForge come from sales and operations teams, and they arrive through ChatGPT rather than through code. This guide is the workflow those teams use.

First, your company template

Do this once, before the first report. Enable the app (search SlideForge in ChatGPT's apps, or open the listing and choose Try in chat), then write:

@SlideForge here is our company template, call it "Sales weekly"

A drop-zone appears in the result card. Drop the company .pptx or .potxthere; the file goes straight to SlideForge, not through the chat. From now on every deck you ask for renders natively on that template — its own slide masters, cover, section dividers and footer — so the report looks like every other deck in your company rather than like a tool's default. Your readers never see a foreign design.

Then, this week's report (about five minutes)

1. Give ChatGPT this week's numbers. Paste them, or upload the spreadsheet export and let ChatGPT read the values. The file stays in ChatGPT; only the values it extracts are sent to render.
2. Start the request with @SlideForgeand describe the deck slide by slide. Name each slide's job, not its design — the example below is enough.
3. Download the .pptx from the result card and read the fidelity report under it: it lists, per field, what was bound as supplied. Then open the deck and check it as you would any deck a colleague built.

What to write

@SlideForge weekly sales report, week 37, on the "Sales weekly"
template, six slides:

1. Cover: "Sales weekly — W37"
2. KPI scoreboard against target: bookings $412k vs $450k,
   new logos 9 vs 10, pipeline created $1.9M vs $1.5M,
   win rate 24% vs 22%
3. Bookings trend, last 8 weeks ($k): 380, 395, 360, 410,
   445, 430, 398, 412 — target line at 450
4. Region table, actual vs target ($k): EMEA 180 / 200,
   NA 165 / 170, APAC 67 / 80
5. Status board: pipeline coverage on track, hiring at risk,
   churn off track — one line each on why
6. Next week's focus: three bullets

Slides render in parallel into one file with page numbers. The engine authors no figures: a number you did not supply is not invented, and the fidelity report names any wording that was completed for you.

The layouts it picks

ChatGPT chooses from a catalog of 200+ maintained layouts, and the report family is the deepest part of it. For a weekly KPI deck you will mostly see:

• Target-progress scoreboard — each KPI with a bar toward its target or quota. Sales attainment, OKR progress.
• Delta scoreboard — the movement since last week is the focal cue, not the absolute value.
• Status board — each metric mapped to on track, at risk or off track against a threshold. Operations, service health, risk.
• Exception board — the off-track metrics take the attention zone; the healthy ones recede.
• Trend line — one measure over the last weeks with the latest value labelled, an optional target line, or a dashed forecast segment.
• Grouped bars and tables — regions, teams or products against plan.

Week two: the same deck, new numbers

Stay in the same ChatGPT conversation and write "@SlideForge same deck as last week, W38 numbers:" followed by this week's figures. ChatGPT sends the same layouts with the new values, on the same template. Because the engine is deterministic — the same layout with the same input renders the same slide, and there is no model in the render path — the only thing that changes is the numbers. The people who read the deck see the format they saw last week.

What else the app does in the same conversation

• Translate the finished deck — "@SlideForge translate this deck to German": 35 languages, layout and formatting preserved, $0.02 per slide.
• Bring last quarter's deck in — drop a PDF of an existing report and get editable slides back, real text boxes and shapes, not a picture.
• A logo on every slide — upload it once; free on every account.
• Cover, dividers and closing from your template — ask for them and the deck uses the template's own cover, divider and closing slides, filled with your title and date.
• Ask what it can draw — "@SlideForge what layouts do you have for a funnel?" browses the catalog for free before anything is billed.

Why not let ChatGPT write the file itself?

ChatGPT can produce a .pptx by writing python-pptx code in its sandbox, and for a one-off that is fine. For a report that carries your numbers to your manager every week, the difference matters: in generated code your figures pass through the model on the way to the slide, the layout varies from run to run, nothing checks the result, and it cannot render on your company's masters. The SlideForge app binds the values ChatGPT sends exactly as sent, renders them into maintained layouts on your template with overflow validation, and reports per field what happened. A slide with a blocking defect is not billed.

When it should run by itself

If the numbers already live in a spreadsheet or a database and the deck should be in inboxes at seven on Monday, the same engine is available without ChatGPT: as an n8n node, a Zapier or Make app, or from a few lines of Python. The spreadsheet rows map onto the same layouts and the same template, and the workflow emails the .pptx.

Frequently asked questions