This shift from Tool to Co-worker is spot on.

Once software starts taking action on its own, the bar for reliability goes way up. A bug in a traditional feature is Annoying, but a mistake by an AI agent feels like a breakdown in trust.

For me, agents haven’t completely replaced core SaaS tools yet mostly because that trust takes time to build but they’ve definitely changed the baseline expectation. If an Agent still needs constant babysitting, it just feels like an automated feature with better marketing. Knowing when to ask for clarification and taking initiative is where the real value lies.