I make slides almost every week. User research summaries, product memos, occasional conference talk. In the past, it would take me hours. Not because I'm slow with PowerPoint, but because I'd get stuck on structure. I knew what I wanted to show, but not what to say first.
The outline came together faster
Last week I had a pile of interview notes from user testing. Normally I'd read them all, then move items around on a virtual whiteboard. This time I pasted the raw notes into an AI chat and asked it to find the top three themes. It returned a rough outline with a recommendation slide. I didn't use it as-is. The deck I delivered was significantly different. But the starting point saved me about forty minutes.
What surprised me is that the AI's suggested order made me rethink the story. It put the "what users did" before "what they said", which I wouldn't have done. I changed it back, but the friction was useful. It made me justify my own logic.
Visuals stopped being a bottleneck
Imagery used to be the most tedious part. I'd search stock photo sites for keywords like "team meeting" and get the same tired results. Or I'd spend time making simple diagrams by hand. Now I can generate a decent placeholder chart from a CSV and ask an AI to find a metaphor for the data. The result is usually not final. But it's good enough to show during a draft.
For a recent internal demo, I wanted to illustrate a two-sided marketplace. I sketched a crude set of boxes and arrows, then described it to an image generator. The output was over-designed and full of text artifacts. But the proportions were right. I ended up rebuilding it in the slide tool myself. The AI saved me from a blank canvas, not the final work.
The real skill is knowing when to ignore AI
I've noticed something about my own workflow. The parts AI helps with are the parts I used to dread. Structure, first drafts, visual placeholders. The parts it doesn't help with are the parts that actually matter: knowing what the audience needs, deciding what to emphasize, and making sure every slide has one point.
While building Huanjian, we've seen similar patterns in how people use AI for slides. Some users ask for a full deck and then strip it down. Others use it to generate talking points when they're stuck. The best decks still come from people who can tell the difference between a suggestion and a fact.
I'm still figuring out where to draw that line. Last month I let an AI write the entire talking track for an investor update. It was phrased cleanly, but it felt hollow when I read it out loud. I rewrote most of it from the notes I had jotted down during a walk. That walk helped more than any tool.
So here's where I've landed. AI changed how I start decks, not how I finish them. It lowered the activation energy for the first ugly draft. I'm ok with that. The slide-making workflow is still mine, but the heavy lifting at the beginning is shared.