Assessment Design for the AI Era: The Teacher Who Can Still Prove What Students Can Do
Why This Field Matters
Anthropic, OpenAI and other labs are distributing educator and student tools free or at deep discounts, moving into a market estimated at $6tn. When the price is effectively zero, adoption runs ahead of whatever the school was ready for.
What breaks first is not instruction. It is assessment. Teachers have always inferred a student’s state from submitted work, and that inference rested on an assumption: someone who can write this essay understands this concept. The tools cut that link. Submission quality goes up while the information a teacher can actually read out of it goes down.
Policy is not reversing either. The conversation has moved from blocking teen access to opening it safely, and opening it comes with an obligation to report outcomes. Usage is also far from uniform: some students never touch these tools, others run every assignment through them. The same worksheet now means different things to different students in the same room.
So what schools need is not a teacher who is good with AI. It is someone who can still make a defensible call about what a student can do while the tools are sitting right there. The role has no standard title yet, but the schools facing a renewal or budget review start looking for it first.
Required Skills
Start with the ability to decompose goals into observable behavior. An objective like “thinks critically” becomes unmeasurable the moment a model is in the loop. It has to come down to something the student must do in the moment: here are two sources that contradict each other, tell me which one you trusted and why.
Second is designing unaided checks. You cannot convert every assignment into a proctored exam, so you build a structure where tool-assisted work is followed by a short verification step. Five minutes of oral questioning, or two variant problems, is often enough. What matters is not length but whether the assignment and the check target the same concept.
Third is reading process evidence. Revision history, chat transcripts, the delta between drafts already exist. Used to catch cheating, this material poisons the classroom; used for diagnosis, it shows things that were never visible before. Being able to explain that distinction is what makes the practice defensible to parents and administrators.
Fourth is documentation inside the institution. Bundle the rubric and the AI-use policy into one document, publish it before the term starts, and settle disputes against it. The teacher who can write that document is the practitioner of this specialization.
Career Path
Start with one subject you already teach. Attach unaided checks to three assignments across a term and record where results diverged from submissions. That record is the only asset the next step requires. When you take the standard to a department or grade-level meeting, the role becomes school-wide.
From there the job is designing assessment policy for a whole school or academy: which tools are permitted how far, which assignments carry a check, what format the results are reported in. The person who prepares the renewal-meeting packet comes from here.
There are exits outside the classroom too. Learning design and assessment roles at edtech companies, district or foundation policy work, and teacher training program development all draw on the same skill. They have one thing in common: instead of banning or celebrating the tools, they need someone who can put in writing how anything gets verified once the tools are a given.
Tags
References
Ready to Start?
Everyone above started just like you. Pick one thing and do it today!