Everyone Hates Making Slides. You're Doing It in the Wrong Order.

AI can build a deck in two minutes, and it'll look generic. The fix is the order you work in: structure first, design second.

Making slides is one of the most universally hated tasks in office work. AI was supposed to fix that. For most people, it didn't.

You've probably tried it. You typed "make me a presentation about X" into ChatGPT or Gamma, and what came back was technically a deck: white slides, generic icons, and bullet points that were really just your prompt chopped into four lines. Usable in the way instant coffee is drinkable. Not something you'd actually present.

The tools aren't the problem. The order is. Most people ask AI to design before they've decided what the deck should say, which is like asking a contractor to paint a house that doesn't have walls yet.

Do it in the right order and a rough doc becomes a finished, presentable deck in about twenty minutes. Here's the order.

Why "make me a presentation" produces garbage

A presentation is two separate things wearing one coat. There's the structure (what each slide says and the order the argument flows in) and there's the design (how it looks). They're different jobs, and AI is good at them in different ways.

When you type "make me a deck about X," you're asking the AI to invent the structure and the design in one shot, from almost nothing. It has no idea what your actual argument is, so it guesses, and the guess is generic. Then it dresses up the generic guess in a template, and you get the white-slides-and-bullets result everyone recognizes.

But the fix is simple: separate the two jobs. Nail the structure first, in plain text, where it's fast to change. Only then hand it to a design tool. This "outline first" discipline is the single thing that separates a good AI deck from the instant-coffee version, and almost nobody does it.

Step 1: Turn your doc into a slide outline

Start with whatever you have. A rough doc, messy notes, a report, even a long email. Paste it in and ask for structure, not slides.

"Here's a rough document. Turn it into a slide-by-slide outline for a presentation to [audience]. For each slide, give me: a clear headline that states the point (not a vague topic), and 2 to 3 supporting bullets in plain language. Aim for [number] slides. The deck should follow one clear narrative arc from problem to resolution. Don't design anything yet, just the structure and content."

Read what comes back as a story, not a slide list. Does it flow? Does each slide earn its place? Is there a point that's missing, or three slides saying the same thing? This is the stage where you fix the thinking, and it's fast because it's just text.

Spend most of your twenty minutes here. The structure is the deck. Everything after is decoration.

Step 2: Force one idea per slide

The most common structural flaw is slides that try to say three things at once. The fix is a single follow-up prompt.

"Go through this outline and make sure each slide makes exactly one point. If any slide is carrying more than one idea, split it. If two slides are making the same point, merge them. A viewer should be able to get each slide's message in three seconds."

One idea per slide is the difference between a deck that lands and a deck the audience reads instead of listens to. This prompt enforces it without you having to police every slide by hand.

Step 3: Generate the actual deck

Now, and only now, you bring the structure into a design tool. You've got three realistic paths.

The fast path is Gamma. It's purpose-built to turn an outline into a designed deck, complete with layout, fonts, and images, in a couple of minutes. The free tier handles up to 10 slides ("cards") per generation and runs on a one-time pot of credits, so it works for a deck or two before you'd upgrade. Paid plans start around $10 a month, which removes the Gamma watermark and gives you enough credits for regular use. As of 2026, Gamma also plugs into Claude directly, so you can generate the deck without leaving the chat where you built the outline. Paste your finished outline in, pick a theme, and you have a real deck.

The content-quality path is Claude or ChatGPT plus a design tool. Claude writes the strongest content and speaker notes of the bunch but doesn't produce designed slides on its own. The standard professional workflow is Claude for the words, Gamma or Canva or Google Slides for the look. Combined, that's roughly $30 to $35 a month, and it's what a lot of consultants actually use.

The built-in path is Claude's own deck output. Claude can generate an HTML or PowerPoint-ready deck through its file skills, which is enough for internal or low-stakes presentations where design polish matters less than just getting it done.

For most people: build the outline in Claude or ChatGPT, then push it into Gamma. Best content, real design, twenty minutes total.

Step 4: Speaker notes, almost for free

The part people forget until the night before. Once the deck is built, the notes are one prompt away.

"Write speaker notes for each slide. For each one, give me three or four sentences I can say out loud that expand on the slide without just reading the bullets. Conversational, like I'm talking to the room, not reciting. Flag any slide where I should pause for a question."

The "expand on the slide without reading the bullets" instruction matters, because reading your slides aloud is the fastest way to lose a room. The notes give you something to actually say.

The part where this makes money

Plenty of people will pay to never make a deck again.

Consultants, coaches, small business owners, anyone who pitches but hates slides. They have the content, a report, a proposal, a set of notes, and no time or taste for turning it into something presentable. The service is simple: they send you the raw material, you run the structure-first workflow, polish the result, and hand back a finished deck.

A deck that takes you twenty-five minutes with this workflow is something people pay anywhere from $50 to several hundred dollars for, depending on the stakes and the polish. Pitch decks and investor decks sit at the high end, because the cost of a bad one is enormous. The same "show, don't pitch" move works here: rebuild a few slides of someone's existing ugly deck, send the before-and-after, let the contrast sell it.

It pairs naturally with the proposal-writing and spreadsheet services, because the same clients who need those usually need decks too.

The honest limits

Three things to keep you honest.

Verify every number. AI presentation tools are notorious for generating plausible statistics that turn out to be invented. If a slide cites a figure, confirm it came from your source material and not the AI's imagination. A made-up stat on a slide in front of a room is a credibility killer.

The first draft is not the final deck. Across the board, these tools are better at producing a strong first draft than a send-it-as-is result. Budget a few minutes to fix the slide where the AI misunderstood your point, swap the image that's slightly off, and cut the slide that doesn't need to exist. The twenty minutes gets you to "almost done," not "done."

Taste is still yours. AI can lay out a slide. It can't tell that your big reveal should be its own slide with five words on it, or that the data belongs before the recommendation, not after. The narrative judgment, the sense of what will actually move this specific audience, is the part you bring. It's also the part that makes the deck yours instead of generic.

If you want the broader system for handing AI the office work you've been grinding through by hand (decks, documents, data, email, planning), I just published Your AI Operating System: The Beginner's Field Guide to Letting AI Do Your Busywork on Gumroad. Volume 1 of my AI for Real Life library. Volume 4 will go deep on AI for office work, where deck-building workflows like this one get the full treatment. Launch price for Volume 1 is $19, and existing buyers get every future volume free as I release them.

The reason your AI decks looked generic was never the AI.

It was that you asked it to decorate before you'd decided what to say. Fix the order, and the slide you used to dread becomes the easy part.