In the last week, I’ve seen some pretty terrible uses of AI and some surprisingly great ones. I’ve had thoughtful conversations with colleagues and friends about AI - what they love and what they don’t.*
AI is a powerful tool with expanding use cases. It isn’t inherently good or bad. There are valid concerns about how models are built and used, and those deserve real attention. But the biggest difference between success and failure (the outcomes) in every example I’ve seen recently has been the same: how involved a human is.
When AI is used to accelerate a task someone is already responsible for, that person can work faster, think more clearly, and produce better outcomes. When it’s used to replace ownership, when someone simply offloads a task they are accountable for, the result is usually obvious. We all recognize it - and we don’t like it.*
You’re still accountable for what you build.
Use AI to accelerate, not abdicate. Human involvement isn’t a limitation, it’s the key to making AI actually work.
- Note: These aren’t em-dashes; I wrote this. 😎