I used to start every deck with a blank slide and a vague sense of dread. My old workflow was simple: copy a previous deck, replace the text, then spend two hours moving boxes around until everything lined up. The writing part was maybe twenty percent of the time. The rest was formatting anxiety.
I've been using AI tools in this workflow for about a year. Not as a magic button that spits out a finished deck. More like a thinking partner that happens to be very fast and never gets bored.
The blank slide problem
The worst moment was always the start. I'd have a topic and a deadline, and nothing else. Opening PowerPoint and staring at the default template made it worse — that big empty canvas asks you to commit to a layout before you've figured out what you want to say.
What changed first was my drafting habit. I stopped opening slide software until I had a text document with the rough storyline. Just one sentence per slide, written in plain text, like I was explaining the topic to a teammate over lunch. That small shift removed most of the pressure. But it created a new problem: turning that rough outline into actual slide content still took forever.
What I do now
Now I hand that rough outline to an AI and ask it to expand each point into slide-level chunks — a short title, one idea, a few bullets. I don't accept the first version. I edit it, push back, rewrite. But having something concrete to react to is much easier than generating everything from nothing.
The second change is structural. While building Huanjian, we found that separating content from layout makes this whole process work better. The AI suggests a structure, the tool applies a layout, and I only adjust the edges. I used to spend twenty minutes aligning boxes. Now that's just not a step anymore.
A friend asked me the other day whether I could still make a deck without AI. I think I could, but I'd be slow and annoyed. The real change is that I no longer treat slide layout as part of the thinking. It's a rendering problem. The thinking happens before the slide exists.
This is the part where I'm supposed to say AI is great for everything. It's not. For data-heavy slides, I still do the numbers myself. And when a deck needs a strong point of view, the AI's first draft is always too polite. It hedges, it balances, it lists pros and cons. For an internal strategy deck, that's useless. The sharp opinions have to come from me.
What still feels like work
Narrative flow is the thing I still wrestle with. A slide deck is not a document. People read it in a nonlinear way, they jump back and forth, they zone out. AI can give me well-structured bullets, but it doesn't know when the audience is going to lose interest. I'd rather have three pointed slides than eight thorough ones, and that instinct is something I can't delegate.
I'm also less sure about what this does to the craft of presenting. When the slides build themselves, the thinking moves elsewhere — into delivery, into Q&A prep, into the spoken narrative. That might be fine. But I've caught myself spending less time on the material than on reacting to what a tool generated. It's comfortable, and comfortable isn't always good.
Honestly, my workflow is still settling. I know the old way was worse. I just don't know yet what this new way is going to cost.