I've been making slides for over a decade. When AI writing tools appeared, I was skeptical. Then I tried them, and now I use them almost every week. But not the way the demo videos show.
Here's the honest version of what I've learned.
The first draft is better than I expected
Last month I had to put together a proposal deck for a potential client. I didn't know where to start. I typed a few sentences of what the deck needed to cover into an AI tool, and within a minute I had a skeleton. Ten slides, each with a title and three bullets. It was... fine. Not great, not terrible.
The thing that surprised me — it wasn't the content. It was the structure. The AI suggested an order I hadn't thought of. A section on risks before the implementation outline. That change made the whole deck more honest. I kept it.
Since then I've started using AI for the blank-page problem. It's genuinely useful when the alternative is staring at an empty screen and wondering where to start. It's not that the output is brilliant. It's that something exists to react to.
Where it falls apart
The problems start when you treat the AI draft as nearly finished. I've learned this the hard way.
Last week, a teammate on my team asked me to review a deck he'd made with an AI slide generator. It looked clean. Same fonts, nice color blocks, icons everywhere. But on slide three, there was a bold claim about our product's performance that wasn't true. The AI had invented a number. It put it in a big chart with a confident gradient fill. Nobody added it on purpose. It just appeared because AI tools fill gaps by making things up.
That's the core issue. AI tools are remarkably good at producing slides that look like a human made them. They are less good at knowing what's true, what matters, and what your actual point is.
I've also noticed the tone problem. AI-generated slides default to corporate dialect — "unlock synergies", "drive engagement", "best-in-class." That language is a red flag to most real audiences. If a slide uses the phrase "leverage" more than once, I rewrite it. If the AI suggests "we're disrupting the ecosystem", I delete it and write what we actually do.
What I do now
My current workflow is a mix. I use AI for the outline and the first rough draft. Then I go through every slide and ask two questions. Is this true? Is this something I would say out loud?
The second question matters more than I expected. Most AI-generated slides are things you would never say in a meeting. They're too polished. They have no personality. A slide that says "Quarter-over-quarter revenue growth demonstrates sustained market adoption" is technically fine. But if I'm presenting, I'd rather say "We grew a bit, but not as much as we'd hoped." Then the slide shows a simple chart.
I've started treating AI like a junior colleague who is fast, confident, and often wrong. You wouldn't let that person send a deck to a client without checking. Same logic applies.
While building Huanjian, we've thought a lot about this. Most AI presentation tools focus on generating slides from scratch. We think that's the wrong default. The better approach is to let AI help you think, then let the human own the final draft. The AI can clean up formatting, turn messy bullet points into readable slides, and handle the tedious work. But the meaning should come from a person.
Anyway, I'm still figuring out the right balance. Some days I let AI do more. Other days I type everything by hand. The one thing I've stopped doing is assuming the AI knows what I'm trying to say. It doesn't. It's just good at appearing like it does.