Building a Pitch Deck in an Afternoon With Copilot
You have a meeting with an investor in four days and a pitch deck that exists only in your head. This is the situation Copilot's PowerPoint integration is actually built for: not polishing a deck you've already drafted, but getting from nothing to a structured, arguable first version fast enough that you still have two days left to fix the parts that matter. If you haven't used Copilot much yet, the Complete Beginner's Guide to Copilot covers the basics this workflow assumes.
The trap most founders fall into is trying to write the deck and design it at the same time. You open PowerPoint, pick a template, and immediately start fussing over whether the title slide font looks credible, before you've even decided what the ten slides are going to argue. Copilot's agent mode works better when you separate those two problems: get the argument structured first, then let Designer handle the visual layer once the content is right.
Start with the pitch in one paragraph, not an outline
Before opening PowerPoint at all, write the one-paragraph version of your pitch: the problem, who has it, why now, what you built, and how big this could get if it works. This is the paragraph you'd say if someone asked "so what do you do?" at a dinner party and you had ninety seconds.
Here's a real one, for a founder building Chairback, a scheduling and booking app for independent hairstylists who rent chairs at salons instead of working as employees:
I'm building Chairback, a scheduling and payments app for independent hairstylists who rent a chair rather than working as salon employees. There are roughly 400,000 of them in the US, and almost all of them currently juggle Instagram DMs, a paper appointment book, and a separate payment app to run their business. Chairback combines booking, deposits, and reminders into one app built specifically for chair renters, not general small businesses. We have 340 stylists using it in three cities after four months, with no paid marketing. I'm raising a $750,000 seed round to hire two engineers and expand to ten cities. Turn this into a 10-slide pitch deck outline for a seed-stage investor meeting: one line per slide describing what that slide needs to prove, in the order a skeptical investor would want to see it.
”This is the single most important prompt in the whole afternoon. Notice what it includes: a specific number (400,000, 340, $750,000), a specific competitor behavior (DMs plus paper plus a separate payment app), and a specific ask. A vague version of the same prompt, "help me make a pitch deck for my scheduling app," gives Copilot nothing to work with except the generic ten-slide template every accelerator's website already has. The output would look like a deck. It wouldn't sound like your business.
Here's a representative version of what actually comes back from that prompt, before any reordering happens:
PowerPoint
The outline Copilot returns from the prompt above, illustrated, not a real screenshot
| Slide | What it needs to prove |
|---|---|
| Title | Chairback exists and what it does, in one line |
| Problem | Independent stylists run their business across DMs, a paper book, and a separate payment app |
| Market | Roughly 400,000 chair-renting stylists in the US, an underserved segment |
| Product | Booking, deposits, and reminders combined into one app built for chair renters specifically |
| Traction | 340 stylists across three cities in four months, with no paid marketing |
| Business model | How Chairback charges and what the unit economics look like |
| Go-to-market | How the next ten cities get reached from here |
| Competition | Why general small-business tools don't fit this specific job |
| Team | Why this founder is positioned to win this market |
| Ask | $750,000 to hire two engineers and expand to ten cities |
That's the raw output. Notice it still puts the market-size slide before traction, the exact ordering problem the next section catches before anything gets built beyond an outline.
Tip
Ask for the outline before you ask for slides. Reviewing ten one-line descriptions takes thirty seconds and catches a wrong argument order before you've built anything on top of it. Reviewing a finished deck to spot the same problem takes much longer, and by then you're emotionally attached to slides you now have to cut.
Let agent mode build the actual slides
Once the outline holds up, either edit it directly or tell Copilot which slide order you want, then have it build the deck. In PowerPoint, agent mode does this as a real multi-step edit: creating slides, writing headline and body text for each one, and applying a layout, not just handing you a block of text to paste in yourself.
- 1
Approve or adjust the outline
Read the ten one-line slide descriptions. Reorder, merge, or cut before anything gets built. For Chairback, the market-size slide and the traction slide needed to swap places, since 340 users with zero marketing spend is the more convincing opener than a market-size number any investor has heard a version of before.
- 2
Generate slide content in batches, not all at once
Ask for three or four slides at a time and read each batch before continuing. A single "build the whole deck" instruction produces content Copilot can't easily revise later without regenerating slides you already liked.
- 3
Push back on generic phrasing immediately
If a slide says something a hundred other decks could say word for word, that's a sign it needs your specific detail added back in, not a sign the tool failed.
- 4
Ask for the appendix separately
Detailed financials, competitive landscape, and team bios belong after the core ten slides, built once the narrative is locked, not interleaved with it.
A real follow-up prompt, after the first batch came back too soft:
The traction slide currently says "Chairback has seen strong early growth and positive user feedback." Replace it with the actual numbers: 340 active stylists across three cities in four months, zero paid marketing spend, and a specific retention detail if I give you one. I'll tell you the retention number next. For now, cut every sentence that isn't backed by a number I've given you.
”That instruction, cut every sentence that isn't backed by a number, is worth reusing on almost any slide agent mode drafts for you. It's the fastest way to strip the generic confidence-building language that reads fine in isolation and terrible next to a competitor's deck full of specifics.
Bring in Designer once the content is locked
Copilot in PowerPoint can generate images with Designer's Image Creator, so you don't need to leave the deck to create a cover image or a supporting graphic. Image generation depends on your plan and, at work, on admin settings, so check that it's available to you. This matters most on two slides: the title slide, which sets the tone before anyone reads a word, and any slide trying to show a concept a bullet list explains badly, like how a chair-rental relationship actually differs from salon employment.
Generate a title slide image for a pitch deck: a clean, modern illustration suggesting an independent stylist working confidently and professionally, not a stock-photo salon scene. Warm, approachable color palette, no text in the image itself, suitable as a background behind a bold title.
”For a concept slide, a simple two-column diagram usually beats an illustration:
Create a simple before/after diagram for a slide: on the left, "Today," a stylist juggling three disconnected tools, shown as three separate icons with a small frustrated visual cue. On the right, "With Chairback," the same three functions combined into one app icon. Minimal, high-contrast, no clutter, readable at a glance from the back of a room.
”Where Designer isn't the right call
If your product has a real screenshot, a real chart from your own data, or a photo of your actual team, use those instead of a generated image on those slides. Investors can usually tell the difference between a real product screenshot and a generated approximation of one, and using the real thing signals you have something to show, not just something to imagine. Save generated visuals for slides that are illustrating a concept, not proving a product exists.
Rough idea
StartOne paragraph: problem, audience, why now, traction, the ask.
Outline
ReviewTen one-line slide descriptions, reordered and cut before anything is built.
Agent-built draft
BuildSlides generated in batches, generic phrasing replaced with your real numbers.
Investor-ready
PolishDesigner visuals on the title and concept slides, real screenshots everywhere else.
What an afternoon actually gets you
Be honest with yourself about what "done" means here. An afternoon with Copilot gets you a structurally sound deck with your real numbers in the right order and visuals that don't look like a template. It does not get you a deck that's been pressure-tested by someone who's heard forty pitches this month and knows which question you'll get asked first. Before the actual meeting, have a person who has sat on the other side of a term sheet read it and tell you where they'd push back. That step doesn't get shorter just because the drafting step got faster.
Write the one-paragraph pitch with real numbers before opening PowerPoint
Get outline approval before generating any slide content
Build in batches so you can revise without regenerating slides you like
Replace generic claims with numbers on every slide that makes a claim
Use generated visuals for concepts, real screenshots for the actual product
Have an experienced person read the deck before the real meeting
Official sources
- Add an image to your presentation with Copilot in PowerPoint
- Get started with Agent Mode in Word, Excel, and PowerPoint
Checked on September 21, 2026. Features, plans and names change often, so the vendor's own pages are the final word.