Six months ago, my team shipped a new version of Huanjian and I realized I hadn't used it for my own presentations. That's not a great look. So I made a rule: every work presentation gets an AI assist. For the next six months I'd write everything through prompts — from the weekly product update to the quarterly planning deck.
I've done at least thirty decks since then. Here are the things I didn't expect.
The first draft gets a lot better
The hardest part of a presentation has always been the blank page. I'd sit in front of the slides, knowing what I needed to say, but everything came out flat. AI changed that. I started typing my messy bullet points into the chat and asking for a story arc. The first draft wasn't great, but it gave me something to push against. Instead of starting from zero, I was editing. And editing is easier.
For example, I had to present a post-mortem on a missed deadline. My initial notes were just accusations and confusion. I asked the AI to organize the facts into a timeline and then suggest three different narratives. The one I used wasn't any of the three — but reading them helped me realize what I actually thought. That's the value.
I also learned to prompt better. Early on I'd write "make me a deck about our Q3 numbers." That produced the kind of slide you've seen a thousand times: title, bar chart, "we had strong growth." Useless. By the third month I was feeding the AI actual data points and my own half-formed opinions. The responses got sharper because my inputs did.
The AI's voice is still a problem
The more polished a draft was, the less it sounded like me. AI loves a well-organized argument. It loves a confident transition. My presentations, on the other hand, tend to be a bit scrappy. I talk through tradeoffs. I admit when I'm unsure.
At one point I used an AI draft almost verbatim for a talk to our engineering team. It read fine, but after the meeting two engineers told me the language felt off. One said, "You never say 'leverage' that often." I checked the transcript. The AI had used "leverage" four times in a single slide. I don't think I've ever used it in my life.
I started using a custom instruction: "Write like a person talking to colleagues over lunch, not like a CEO giving a press release." That helped. But I also found myself rewriting more by the end of the six months, not less. The AI's first draft was a useful scaffold, but my own voice needed to do the actual building.
It changed my thinking process
The unexpected shift happened when I stopped treating AI as a writing tool and started treating it as a thinking partner. I began using prompts like "here's my core message, here's who's in the room, what's the strongest argument against this?" Or "I have ten minutes, what should I cut?"
That's where the tool finally felt like mine. It didn't just give me content, it made me articulate my own assumptions. I remember putting together an internal pitch for a new feature. I had a deck ready, but before the meeting I asked the AI to poke holes in it. It gave me six objections. Five were silly, but one was real. I changed the deck because of that. I would not have found that weakness without the prompt.
While building Huanjian we found that people don't want the AI to take over the whole presentation. They want it to get them unstuck, and then get out of the way. My experience matches that. The best presentations I made in these six months were the ones where AI did the heavy lifting of organizing raw material, and I did the messy, human work of deciding what actually mattered.
Am I going to keep using AI for every presentation? Probably not for every single one. Sometimes I just open a blank deck and type at empty slides because that's how I think. But I'm not going back to writing from scratch on a regular basis. That version of me didn't have a second pair of eyes.
I'm still figuring out when to trust the machine and when to override it. Someone told me last week that using AI to make presentations is kind of like hiring a ghostwriter who's very smart but has no taste. I'd say the taste part is getting better, but it needs a partner. That's the part that won't change.