Why AI Keynotes Are Broken (And What Audiences Actually Need)
After attending dozens of AI keynotes and delivering my own, I've identified three patterns that make most of them useless — and what I try to do differently.
I've sat through a lot of AI keynotes. Before I started delivering my own, I attended probably 40 or 50 of them across the conference circuit. And I started noticing patterns that troubled me — patterns that I think explain why most people leave AI keynotes without anything actionable.
Pattern 1: The Hype-Terror Pendulum. Most AI keynotes oscillate between utopian enthusiasm and apocalyptic fear. The speaker spends 20 minutes telling you AI is going to revolutionize everything and then 10 minutes telling you it might destroy everything. The audience leaves energized and terrified in equal measure, which mostly produces paralysis.
Pattern 2: The Demo Trap. Speakers spend 30 minutes showing you impressive demos of AI doing impressive things. The audience oohs and aahs. Nobody learns anything, because the demo doesn't transfer — it just creates the feeling of understanding without the substance.
Pattern 3: The Tool Focus. The entire keynote is about specific tools, vendors, or models. Which is useless in 18 months when those tools have changed and the audience has learned nothing transferable.
What I try to do instead: Start with stakes, not with features. Give frameworks, not demos. Be honest about what doesn't work, not just what does. End with one thing a person could actually do differently on Monday morning.