Over the past ten days, we've assessed nine new roles across education, management, and hospitality. One pattern jumped out: education professionals—teachers, guidance counselors, activity developers—came in at 23% effective automation exposure, noticeably lower than our all-time average of 28%. On the surface, this shouldn't be surprising. Teaching involves lesson plans, feedback, scheduling, and documentation; all of it digital-ready. Yet the scores tell a different story.
The gap exists because automation potential and automation reality are different things. Yes, a language model can draft a lesson plan or summarize a student's progress. But teaching lives in relationships. A guidance counselor's value isn't just knowing what to say—it's being someone a student trusts enough to be honest with. A teacher's effectiveness isn't measured only by content delivery; it's built through presence, consistency, and the informal knowledge that accumulates by watching a specific student struggle with a specific concept over weeks. Remove the person, and you remove the foundation that made the tool useful in the first place. That's what shows up in the lower scores.
Worth noting: this dataset is still young—nine new assessments in a week—and mostly seeded from synthetic profiles. Real signal is just beginning. We also haven't seen AI tools already in use for any of these roles yet, which might change as adoption spreads. But the early pattern is worth holding lightly: roles where the relationship is the work tend to have higher human moats than their tooling suggests.
If you're in a role where people come to you partly for who you are, not just what you know, how much of your actual value would survive if a system could do your tasks but not your presence?