The pitch you’ll hear: AI makes evaluation objective. It’s not that simple, and pretending otherwise insults a smart audience.
The point worth examining
AI trained on past decisions can inherit the same affinity and pattern-matching biases humans have — the “founder who looks the part” problem, at scale and faster. That’s the real risk, and it’s worth naming.
But there’s an honest upside too. A well-built system applies the same lens to deck #3 and deck #33. It doesn’t get tired, hungry, or impressed by a logo. It can make bias measurable — something a room full of gut calls never can.
What gets missed
So the truthful claim isn’t “AI is unbiased.” It’s “AI can make bias visible and consistent enough to actually manage.” That’s a smaller promise. It’s also a real one.
““AI removes bias” is a comforting lie. Here’s the honest version.”
— ai.STARTUPJURY field note
Which worries you more — human bias you can’t measure, or machine bias you can?