‎I think the biggest shift is from AI as a feature to AI as the worker. Instead of building another tool that helps someone do a task, I’d look for a task AI can reliably take over from start to finish.
‎

For me, the interesting part is that the first 100 users usually don’t come from some fancy growth strategy. It’s often a few conversations, a couple of referrals, and people genuinely finding the product useful. Once you know why those early users stayed, growth becomes a lot easier to think about.

‎I think the real test comes after the first few months. A demo can make almost any AI support tool look impressive, but real customers are messy. I’d care more about how it handles weird questions, escalations, and keeping context when a conversation moves from AI to a human.