Which education workflows change first
In education, the first workflows changing are the ones teachers do after students go home: drafting lesson plans and worksheets, differentiating one material into several reading levels, writing first drafts of parent communications and report comments, and handling administrative paperwork. First-pass feedback on objective exercises is also shifting. These are preparation and paperwork tasks, and surveys of educators consistently identify them as the largest time drains outside actual teaching.
The classroom itself changes more slowly and less predictably. Students are already using AI, which is forcing changes to assessment design faster than any official adoption plan: take-home essays are being rethought, and in-class demonstration of understanding is regaining weight. The near-term picture is not machines teaching classes. It is teachers with lighter preparation loads working in classrooms where checking authentic student understanding has become a harder and more important task.
Where human accountability concentrates in education
A teacher's accountability was never mainly about delivering content. It concentrates in safeguarding students, managing a room of unequal needs, deciding when a struggling child needs intervention rather than more practice, and assigning grades that follow students for years. These decisions carry legal and ethical weight that no district hands to software. AI can surface a pattern in a student's work, but interpreting it against everything else known about that child stays with the adult in the room.
Motivation and trust are the other concentration points. Getting a discouraged student to try again, handling a conflict between classmates, and making a parent feel heard are tasks built on relationships that accumulate over months. The ILO's framing of generative AI as transforming tasks rather than displacing occupations fits teaching well: the transformable part is preparation and paperwork, and the concentrated part is exactly the human work that made people become teachers.
What educators can build in 90 days
In the first month, use AI only on your preparation load: draft one unit's materials, differentiate a text for three reading levels, and outline parent communications, editing everything before it reaches anyone. Learn your school's policy on student data, and never put identifiable student information into unapproved tools. In the second month, redesign one assessment assuming students have AI access, shifting weight toward in-class reasoning, oral explanation, or process evidence.
In the third month, make your practice visible and shared. Write a one-page AI policy for your own classroom, discuss it with students, and compare notes with colleagues about what saved time and what produced errors. If you work in administration or instructional design, pilot one workflow end to end and document the checking steps it needed. Ninety days will not settle the technology, but it can leave you with tested routines instead of anxiety.
