“I’m not sure what that is, AI added it.”
Heard that in a meeting recently?
At a conference I attended 15 years ago, a prominent researcher had to miss their talk at the last minute due to an illness. A colleague stepped in and did the presentation live, in front of a few hundred people, without having ever seen the slides before. We all understood it was an emergency situation, and the presenter kept it light and made it one of the most entertaining talks I’ve seen. At one point he came to a slide and just said “no idea what this is about, let’s go to the next one” and the room laughed it off.
It’s stuck with me because at the time, the idea that the presenter wouldn’t know more about the work they’re sharing than everyone else in the room was almost absurd. The act of prepping slides for a major conference talk was a forcing mechanism that tested if you really understood your material, and it would be hard to make it in front of the room without being able to explain it and survive tough questions. I can’t think of another time, other than this one emergency situation, that I saw someone present something they’d never seen before.
That is, until the rise of AI turned this from a once a decade event into a routine occurrence. Staring at a blank page trying to craft a narrative used to force us to do the hard work of synthesizing our material into a coherent story. Now it’s easy, and fast, to productionalize a half-formed (or less!) idea into a crisp deck. Add in how AI advances are translating into more and more pressure to accelerate our workflows, and it can quickly become tempting to turn stray thoughts into a one-pager before we’ve worked through the gaps.
AI’s “personality” only makes this harder. Chats with AI can lean toward strong validation, meaning we have to stay vigilant against the natural human tendency to prioritize information that conforms to pre-existing beliefs (called “confirmation bias;” see my prior discussion of this here). I also find that AI can be like a masterful improv comedy partner, always answering with a “yes, and…,” following up validation with an offer to “turn that into a one-pager for your upcoming 1-on-1”.
I don’t recall ever seeing a response like “would you like to pause and think critically about this for a moment before I draft the executive summary?”
Developing work and building the story used to be more distinct phases, and each phase exposed gaps in the other. AI is collapsing them into one as it drives more and more pressure to accelerate our workflows. This means we have to be intentional about checking our own work and iterating on it.
For me, the way to bring back that forcing mechanism is the practice run. Tell the story out loud, even if it’s just to yourself, and even if it’s just an hour before you share it with your audience. You’ll very quickly find the points where you get stuck, where something doesn’t make sense, or where you want to jump four slides ahead rather than follow the flow AI built.
The pressure to accelerate is growing, and the tools are there to do it. But we still own our work. After all, we’re only gaining efficiency if the final product is as good as it used to be. Otherwise the outcome is still “I’m not sure what that is, AI added it.” We just get there faster.
