Is “AI coworker” actually the next interface paradigm?

Lately I’ve noticed more AI products moving away from “chat box as tool” and toward “AI as coworker.” Helio, Slock, Multica and a few others all seem to be playing with this idea in different ways. ​ The interesting part to me is not whether the metaphor is accurate. It is what the metaphor does to user behavior. If something feels like a tool, I expect to open it, use it, close it, and move on. If something feels like a coworker, I expect continuity, responsibility, memory, and maybe even initiative. ​ That changes the psychology of trust pretty fast. A persistent AI coworker might reduce friction because it is always in the workstream, but it might also make people over-delegate before the system has earned that trust. ​ Would you rather have an AI tool you open when needed, or an AI coworker that keeps running in the background?