How to Make a Presentation With AI in Minutes

Staring at a blank slide is one of the most reliable ways to lose an afternoon. You know roughly what you want to say, but turning that into an outline, then into slides, then into something you can actually present is a slog. The good news is that you can now make a presentation with AI in minutes, getting a complete first draft outline, slide copy, and speaker notes from a single well-written prompt, then spending your time on the part that matters: the delivery.

I build pitch decks, training slides, and proposal presentations this way regularly. In this guide I will show you the exact sequence I use, give you copy-paste prompts, walk through a real before-and-after, and be honest about where AI helps and where it absolutely does not (spoiler: the design and the facts are still on you).

How to make a presentation with AI: the right order

The mistake most people make is asking for finished slides immediately. That gets you a generic deck you then have to unpick. The reliable order is: brief, then outline, then copy, then notes, then design. Each step builds on an approved version of the last, so you catch problems while they are still cheap to fix.

I use ChatGPT and Claude for the writing, and a slide tool (Google Slides, PowerPoint, Canva, or Gamma) for the layout. Some tools now generate a whole deck from a prompt directly, which I will cover at the end. If you are choosing a writing tool, I compared them in ChatGPT vs Claude for business tasks.

Step one: brief the AI properly

The quality of your deck is decided here. A vague request gets a vague deck. Use a prompt like this:

Help me build a presentation.
Topic: [your topic]
Audience: [who they are and what they care about]
Length: [e.g. 10 minutes, about 8 slides]
Goal: the one thing I want them to remember or do
Tone: [e.g. confident but plain, no jargon]
Start by giving me a slide-by-slide outline only. No slide text yet.

That last instruction is the secret. By forcing an outline first, you get to fix the structure before a single word of copy is written.

Step two: approve and refine the outline

You will get something like: title, the problem, why it matters, your solution, proof, the ask. Read it as a story. Does it flow? Is anything missing or out of order? Reply with edits in plain language:

Good, but move the proof slide before the solution, drop the slide about history, and add a slide near the end with the three next steps. Then show me the revised outline.

Reordering an outline takes seconds. Reordering ten finished slides with copy and design takes ages. This is why the outline-first habit saves so much time.

Step three: write tight slide copy

Once the outline is right, turn it into actual slide text, with a strict limit:

Now write the slides from this outline.
For each slide give me:
- a short headline (max 8 words)
- 3 to 4 bullets (max 10 words each)
No paragraphs. Slides are signposts, not a script.

The word limits matter. The most common slide crime is cramming a paragraph onto each slide and then reading it aloud. Telling the AI to keep it short forces the kind of lean slide that actually supports a talk instead of competing with it.

A real before-and-after

Here is a concrete example from a client who had to pitch a new service to a board with two days' notice.

Before: He opened PowerPoint, typed a title, and froze. Two hours later he had four overcrowded slides and a knot in his stomach. He kept rewriting the same opening slide instead of building the talk.

After: He pasted the brief prompt above with his topic and audience. In two minutes he had an eight-slide outline. He reordered two slides, approved it, and asked for the copy and speaker notes. Twenty minutes later he had a complete draft deck. He spent the remaining time on what actually mattered: rehearsing out loud and fixing the design. The deck that had felt impossible was done before lunch.

Step four: generate speaker notes

This is the step people forget, and it is the one that makes you look prepared. Ask:

For each slide, write short speaker notes: what I should say out loud, in a natural spoken voice, about 3 sentences per slide. Do not just repeat the bullets.

Now edit them so they sound like you, not like an AI. Read them aloud once. The difference between someone reading bullets off a screen and someone talking to the room with notes for backup is enormous, and this step gets you there fast.

Step five: move it into slides and polish

You now have all the words. Two ways to get them onto slides:

Either way, the design and the facts are now your job. Fix the spacing, match your colors and fonts, swap in real images, and check every number, name, and claim. AI-generated decks love to insert plausible statistics that are simply wrong, so verify anything factual before you present it.

The caveats: read this part

Design is not solved. AI gives you words and a rough structure. It does not give you a tasteful, on-brand layout. Auto-generated decks often look generic or slightly off. Budget time to make it look like yours.

Facts need checking. Any statistic, quote, or claim the AI produces could be invented. A wrong number in a board deck is worse than no number. Verify everything factual against a real source.

It will sound generic unless you push. The first draft is competent and bland. Your edits, your voice in the speaker notes, and your real examples are what make it land. Use the draft as a starting point, not a finished product. Better prompts help a lot here, which I cover in how to write good AI prompts for business.

Privacy. Do not paste confidential figures, client names, or unreleased plans into a consumer chat tool if you would not want them stored on someone else's servers. Anonymize sensitive numbers first. I cover where the line sits in is it safe to upload business data to ChatGPT.

When to do it by hand vs automate it

Building a one-off presentation in a chat window is exactly the right move. It is fast, it removes the blank-page paralysis, and you stay in control of the message. For most people, most of the time, that is the whole answer.

But if you produce the same kind of deck on a regular schedule, say a weekly status update or a monthly report presentation built from the same data every time, that repetition is a signal. At that point a small automation can pull the latest numbers, generate the slide copy, and assemble a draft deck on a schedule with no chat window at all. That is where a manual habit becomes a system, and I wrote about exactly that line in when to stop doing it manually and automate it and in business automation for small business.

If you are rebuilding the same presentation every week from the same source, it is worth automating. I am happy to look at yours and tell you honestly whether it is worth it. You can book a call or reach me through the contact form, no pressure.