Back to Guides
Product ManagersEntrepreneurs

Building a Slide Deck With Copilot in PowerPoint From a Rough Outline

The fastest way to get a bad deck out of Copilot in PowerPoint is to paste in a bullet list and ask it to "make this into slides." You'll get slides. They'll also look like every other auto-generated deck: one idea per slide regardless of how much that idea actually needs, generic stock-photo-style graphics, and a structure that mirrors your bullet list's order instead of the order an audience would actually want to hear it in. The fix isn't a cleverer one-line prompt. It's treating deck-building as the same outline-then-build sequence you'd use for any other document, just applied to slides instead of paragraphs.

Note

This assumes some familiarity with Copilot in PowerPoint already. If you're starting from zero, the Complete Beginner's Guide to Copilot covers the basics. And since building a deck end to end leans heavily on multi-step editing, Copilot's agent mode overview is useful background for what's actually happening behind a prompt like "restructure this deck."

The outline: real, and genuinely rough

Here's the starting material for this walkthrough, a founder's raw notes ahead of a board update, exactly as messy as these usually are in real life:

  • revenue up, mostly from the enterprise tier
  • churn still an issue in the SMB segment, we think it's onboarding
  • hired two engineers, one left
  • runway is 14 months at current burn
  • need board's input on whether to raise now or wait
  • new pricing page launched last month, too early to tell if it's working

That's a list, not a deck. It has no order an audience should hear it in, no sense of which point is the headline and which is a footnote, and no obvious slide-per-point mapping, since some of those bullets are a whole slide's worth of content and others are a single supporting detail.

Step one: ask for a slide structure, not slides

The prompt that asks for structure is doing real work here, and the gap between a weak version of it and a strong one shows up directly in the slide order you get back, not just in how the words are phrased.

Bad

One bullet, one slide

"Turn these notes into slides." returns six slides in the exact order of the original list: revenue, churn, hiring, runway, the raise question, and the pricing page, each one just the bullet restated as a headline.

Better

Grouped, still misordered

"Group these into themed slides for a board audience." merges churn and pricing into one slide and folds hiring into another, four slides total, but still opens with revenue and pushes the raise decision to second-to-last.

Excellent

Ordered for the audience

Naming the board's actual priority explicitly moves the runway and raise decision to slide two and turns hiring into a single supporting line instead of its own slide.

Bad prompt: "Turn these notes into slides."

That produces something like: six slides, one per bullet, in the exact order the notes were written: Revenue, Churn, Hiring, Runway, The Raise Question, Pricing Page. It's technically a deck. It's also organized by what order the founder happened to jot the notes down in, not by what a board needs to hear first.

Better prompt: "Group these into themed slides for a board audience."

That's closer: churn and the pricing page merge into one slide since they're related, and hiring loses its own dedicated slide. But the result still opens with revenue growth and doesn't move the raise decision earlier than slide four, because "for a board audience" tells Copilot who's reading without telling it what that audience is actually there to decide.

Excellent prompt:

Prompt
“

I'm building a board update deck from these rough notes: [paste the bullet list above]. Before generating any slides, propose a structure: how many slides, what each slide's headline should be, and what order they should go in for a board audience that cares most about runway and the raise decision. Don't write slide content yet, just the structure.

”

Copilot in PowerPoint often does part of this on its own: when you start a new presentation it may ask who the audience is and what style you want, and propose an outline you can refine in chat before any slides are generated. Asking for structure explicitly, with your own priorities named, solves the "one bullet equals one slide" problem more directly. A board audience needs the runway and fundraising decision framed early, not buried after two slides about hiring. Reviewing the proposed order before any slide exists means you can move the fundraising question to slide two instead of discovering, after the whole deck is built, that it landed on slide seven.

PowerPoint

Proposed slide structure, illustrated

Slide 1, Title: Board Update, Q3. Slide 2, Runway and the Raise Decision: the open question the board needs to weigh in on. Slide 3, Revenue Growth: driven by the enterprise tier. Slide 4, Churn, SMB vs. Enterprise: the segment gap and the onboarding theory behind it. Slide 5, Team and Product Update: the hiring change and the new pricing page, as one combined status slide.

  1. 1

    Push back on the ordering, not just the wording

    If Copilot's proposed structure leads with hiring and ends with the fundraising decision, say so directly: "Move the runway and fundraising question earlier, that's what the board actually needs to weigh in on." This is the cheapest point to fix a structural problem, before any slide text or design exists.

  2. 2

    Decide what gets its own slide versus a supporting bullet

    A single engineer leaving probably doesn't need its own slide; it's a supporting detail on a hiring slide. Tell Copilot explicitly which points should be combined, so it doesn't default to a rigid one-bullet, one-slide mapping.

  3. 3

    Approve the final structure before moving on

    Once the order and slide count look right, say so clearly, so the next request can build directly on an approved plan instead of Copilot guessing whether the structure is final.

Step two: let agent mode build the deck from the approved structure

With the structure agreed, ask for the actual slides in one request, since agent mode can build several slides worth of content and formatting in one pass rather than one slide at a time.

Prompt
“

Build the deck using the approved structure. For the runway and fundraising slide, phrase it as an open question for the board, not a recommendation, since we haven't decided yet. Use our standard title slide and section header styling throughout. Keep body slides to no more than 4 lines of text each.

”

Two details in that prompt do real work. "Phrase it as an open question, not a recommendation" stops Copilot from writing a confident, decisive line like "We recommend raising in Q2" on a slide about a decision that genuinely hasn't been made yet, which is a subtle but real way generated content can misrepresent where things actually stand. "No more than 4 lines of text each" heads off the generic-deck look where every slide is a dense paragraph broken into bullet points; a tight line limit forces the content to actually be slide-shaped instead of document-shaped.

Here's a representative version of what that produces for the highest-stakes slide in the deck:

PowerPoint

Slide 2 content, illustrated

Title: Runway and the Raise Decision. Fourteen months of runway at the current burn rate. Two paths in front of the board: raise now, while enterprise growth makes the story strong, or wait for more renewal data before going to market. Board input needed on which path, and by when.

Notice what didn't happen: no line on that slide says "we recommend raising now" or gives a specific target valuation, because the prompt explicitly asked for an open question, not a stance the company hasn't actually taken. A version of this slide built from a vaguer prompt tends to drift toward sounding more decided than the founder actually is, since a confident declarative sentence is a more natural default output than a genuinely open one.

Step three: add the one visual that needs to earn its place

Rather than asking for a generic supporting image on every slide, add one where it actually clarifies something.

Prompt
“

On the churn slide, add a simple graphic showing the difference in churn rate between SMB and Enterprise segments as two contrasting shapes, not a literal bar chart with numbers, since we don't want to commit to the exact figure on this slide yet. Keep it visually simple, matching a clean corporate deck style.

”

Image generation in PowerPoint is powered by Designer's image tools, and whether you can use it depends on your plan and your admin's settings. Slides with AI-generated images note that in the speaker notes. If it isn't available to you, ask for a native shape-based graphic or a stock image instead. This is a case where a visual genuinely earns its place: it makes the size of the gap between two segments legible at a glance, which a bullet point reading "SMB churn is higher than enterprise churn" doesn't convey nearly as well. It's also deliberately not a chart with exact numbers, since the founder doesn't want to commit an unverified figure to a board slide.

Where generic decks go wrong

The tell that a deck was rushed through in one prompt

A deck where every slide has the same rhythm, headline, three bullets, generic graphic, regardless of whether that slide is announcing a decision point or a minor update, reads as auto-generated even when every sentence is factually accurate. Vary the format: a decision slide can be a single bold question. A quick status update can be one line. Not every slide needs the same shape.

The other common mistake is asking for the whole deck in one prompt without the structure step first. It's faster in the moment and almost always produces something you end up substantially reordering anyway, at which point you've spent the same total time you would have spent approving a structure up front, plus the frustration of untangling an approved-feeling deck that turns out to need surgery.

A final pass worth doing by hand

Once the deck is built, read it once from the audience's seat, not the writer's. For a board deck specifically, that means checking that anything phrased as a decision the board needs to make actually reads as an open question, and that anything Copilot phrased with more confidence than you actually have, a churn cause you suspect but haven't confirmed, a timeline you're hoping for rather than committed to, gets walked back to match what you actually know. Copilot is good at making a deck read as confident and finished. Whether that confidence is earned is still a judgment call for the person presenting it.

Official sources

Checked on September 21, 2026. Features, plans and names change often, so the vendor's own pages are the final word.

Related Guides