For years, opening PowerPoint was the most anxiety-inducing part of my job. The blank slide stared back at me like an empty Word document, except worse — because I also had to think about layout, hierarchy, and visual weight. I used to spend at least an hour fiddling with placeholder text and bullet points before I had anything resembling a narrative.
That's mostly gone now. The first draft is no longer the hard part. The hard part is deciding what I actually want to say before I let the AI take a swing at it.
The Blank Canvas Trap
My old workflow had a specific failure point: the 2 a.m. start. I'd sit down, open a template, and try to structure my thoughts on the fly. The result was always a tangled mess. I'd write the title of slide five, realize it didn't fit the argument, then backtrack to slide two and rewrite everything. It was inefficient, and frankly, it felt bad.
Now I do the opposite. I dump raw notes into a text file first — bullet points, fragments, sometimes just feelings. I don't worry about order or phrasing. Then I paste that mess into an AI assistant and ask for an outline.
That shift alone changed my output quality. I'm not writing from a blank canvas anymore. I'm editing a first draft that took about four seconds to produce. The outline isn't perfect — it usually has a few structural gaps and one or two points that feel irrelevant to the core ask. But it gives me something to react to. That's the whole game. Reacting is easier than creating.
My teammate told me last week that he used this approach for a weekly project update. He gave the AI a list of thirteen status points, asked it to group them by theme, and got back a slide structure that needed maybe five minutes of rearranging. He used to budget an entire evening for that deck. Now he budgets an hour.
The Real Change: Thinking Before Generating
But honestly, the bigger shift isn't the tooling. It's what using the tool forces me to do before I open it. I can't hand an AI a vague idea like "make a deck about the product" and expect anything useful. I've tried. The output looks like a brochure and reads like a buzzword generator.
What works is a short brief. A few lines that say:
- Who the audience is — investors, internal team, customers
- What I want them to feel or do after the last slide
- The three or four points I refuse to cut
- One example or story that must be included
That brief takes me fifteen minutes to write. It used to feel like wasted time. Now it's the most valuable part of the process. I'm not generating slides; I'm clarifying my own thinking. The AI just handles the transcription of that intent into a visual structure.
There was a deck I made last quarter for a feature update. I had a mountain of data about user behavior, but I was stuck on the narrative. The old me would've started writing slides about the chart type I liked best. Instead, I wrote a one-paragraph brief that said, "Audience is existing customers. They need to understand why this change is safe and how it saves them time. The core point is that onboarding time drops by 40%. Use the beta tester quote." The AI took that brief and gave me a five-slide sequence that was nearly ready to go. I spent the remaining time on polish and a better transition between slides three and four.
Garbage In, Garbage Out (Still)
I've seen people complain that AI-generated slides look generic. My take is that they're usually generic because the input was generic. If you ask for "a creative presentation about our new feature," you deserve the stock-photo result you're going to get. The tool isn't thinking; it's pattern-matching. Your job is to inject the specific.
One morning I tried to skip the brief. I was late for a meeting and thought I'd save time by pasting a dense, unedited document into the generator. The output was a shapeless blob of text with no visual rhythm. A wall of bullets. I remember staring at it and feeling the old frustration creep back. The problem wasn't the AI. It was my lazy input.
While building Huanjian, we've run into this exact problem with our own users. People paste vague requirements and expect a masterpiece. The ones who get the most value are those who treat the tool like a sharp assistant who needs clear instructions. The output is only as good as the intent behind it.
I've since learned to prompt in stages. First, I generate just a narrative outline and argue with it. Then I ask for a slide-by-slide breakdown where each slide has a single goal. Only after I'm happy with the text structure do I worry about design. That sequencing matters. It forces me to separate content thinking from visual thinking.
I still don't trust AI to design the whole deck unsupervised. The color choices sometimes clash, and the font pairings can be off. But it gives me a scaffold. I end up moving blocks around instead of building from zero. And for someone who used to loathe the blank page, that's a genuinely different way to work.
I'm still figuring out where the line is between my judgment and the machine's suggestion. There are days when I overthink a slide title and realize the AI's first attempt was fine. Other days, I go back to the brief and realize I didn't actually know what I wanted in the first place.
That's the honest version. The tool didn't make me a better designer. It made me better at deciding what I want to say — which, in the end, was the skill I was missing all along.