We Critiqued 3 "Agentic" Productivity Apps. One Felt Human.

Three productivity apps this quarter lead their marketing with the word agent: one for tasks, one for calendars, one for inbox management. We ran all three through the seven-dimension scale, weighting trust and control more heavily than usual, since an agent that manages your schedule or your inbox is an agent with a lot of room to quietly cause a problem.

The core question we kept asking of each one: when the agent takes an action, can the user see what happened and undo it easily if it guessed wrong. The calendar app passed this cleanly, logging every change it made in plain language and offering a one-tap revert. The inbox app was the weakest here, silently archiving messages it judged low-priority with no visible log and no easy way to find what had been moved.

The task app landed in between, but it earned the standout line in this critique for a smaller reason: its agent explained its reasoning in the same casual tone a helpful coworker would use, rather than a clinical status report. "Moved this to next week since you have three things due Friday already" reads differently than "Task rescheduled: priority conflict detected," even though both convey the same fact. That difference is the whole reason it felt human where the others felt automated.

The principle worth taking from this: agentic does not mean invisible. The apps that earned trust this round were the ones that treated visible reasoning as a feature in its own right, not an afterthought bolted onto a working system. If you are building an agent feature, budget real design time for how it explains itself, not just for what it is capable of doing.