Six months ago I made a rule for myself: no deck goes out without going through AI first. Not "AI helps me brainstorm sometimes" — every one. Internal standups, the quarterly review, a partner pitch, a 20-minute talk I got handed with two days' notice.
I work on presentation software, so this wasn't neutral research. I wanted to know what it actually feels like to depend on it.
Month one was embarrassing
The first few decks looked like a consulting template had a baby with a LinkedIn post. Lots of "three key pillars." Lots of bullets that just restated the heading. I sent one to my manager and he asked, gently, whether I'd been rushed. I'd spent two hours on it.
The mistake was asking the model to think. It's not good at that. What it's genuinely good at is taking thinking you've already done and giving it a shape.
The partner pitch was the low point. I fed in our meeting notes and asked for a narrative. What came back was smooth and completely weightless — it had turned a real disagreement about pricing into a tidy list of "considerations." I rewrote the whole thing at 11pm.
So I changed the input. Instead of typing "make me a deck about Q3 retention," I started talking. Forty minutes rambling into a voice memo while walking the dog, then feeding the transcript in and asking for an outline. The outline came back maybe 70% right. Fixing the other 30% took fifteen minutes. Staring at a blank outline used to take me ninety.
The middle months got weird
Around month three I noticed something I didn't like. I had to sketch an outline by hand on a flight, no wifi, and it was mush. Disconnected points, no spine. Six months earlier I could do that in my sleep.
Is that atrophy, or did I just get used to starting from something instead of nothing? Honestly I'm not sure. I've started doing the flight exercise on purpose, just to keep the muscle.
There was a second problem: sameness. Three decks I made in month two looked like siblings. Same structure, same rhythm, same instinct for where the "so what" slide goes. A teammate told me my roadmap deck felt like it had been written by a committee. He wasn't wrong, and I'd mostly written that one myself.
While building Huanjian we kept running into the same wall from the other side. People don't actually want a machine that writes decks. They want something that gets them from a fuzzy idea to a shape they can argue with. The arguing is the work.
What six months actually changed
The deck stopped being the deliverable. That's the real shift. When structure takes twenty minutes instead of two hours, you spend the difference on the story and on rehearsing out loud. My reviews got better. Not because the