AI assessment redesign is already happening in K-12 schools. It just isn’t happening as a system. It’s happening one classroom at a time, one teacher at a time, and often alone.
That’s the part of this week’s news I can’t stop thinking about. Teachers are doing smart, creative work to find out what students actually know. But when every teacher builds their own answer, students get a different definition of learning in every room.
That isn’t a teacher problem. It’s a leadership gap.
Key takeaways
- Teachers across the country are redesigning assessment for AI on their own: oral exams, more in-class writing, fewer take-home tasks.
- Students are already using AI daily for schoolwork, so the question is no longer whether they use it but what counts as evidence of their thinking.
- When each teacher solves the problem alone, students experience inconsistent expectations. Leaders need shared thinking, not identical practice.
- The best assessment redesign asks students to explain, defend and transfer their thinking, which is exactly what AI can’t fake for them.
What’s happening
On October 28, Reuters reporter Helen Coster published a look at how middle and high school teachers are remaking classrooms for the AI age, based on interviews with a dozen teachers and education experts. The pattern is clear: less take-home work, more in-class writing and discussion, and assignments designed to show how students think, not just what answer they reach.
The examples are worth reading closely. A sixth-grade humanities teacher in San Francisco, Tristen Chang, stopped assigning take-home essays and now uses oral exams, asking students questions about character, plot, theme and metaphor. A New York high school math teacher, Bobson Wong, assigns fewer homework problems and more in-class practice, and now asks students which problem-solving method they prefer and why, since AI can show every step of a solution. A suburban New York English teacher, Meghan Hyland, has made classwork a larger share of grades and told Reuters, “My main ability to understand whether they know how to do critical thinking has to be done in school.” A San Francisco middle school English teacher, Rusty Shabe, spent a month on a single essay with seventh-graders, giving more class time to writing. Others are writing prompts they believe AI handles poorly, limiting laptops, or, like social studies teacher Michael Zagari, using AI to adjust reading levels and encouraging students to use it to start research.
Then there is the sentence that matters most for leaders: school systems remain divided over AI, “leaving many teachers to navigate the transition largely on their own.”
Two days later, NPR’s Rhitu Chatterjee reported on two new surveys showing how common chatbot use has become among students. In a JED Foundation survey of 5,500 students ages 9 to 17, 90% had used an AI chatbot, and nearly two-thirds use chatbots for schoolwork as well as for companionship or when they feel stressed or overwhelmed. Common Sense Media found that nearly half of 9- to 12-year-olds have used an AI chatbot or companion. One student described having chatbots generate sample test questions that he then answers on his own. Another runs her essay drafts through Grammarly or ChatGPT for feedback, then edits.
And in New York City, where AI is currently prohibited for 2-K through eighth grade this school year, NY1 reported on October 28 that students and advocates asked the city to extend its AI moratorium to high schools. One Hunter College High School student said classmates rely on AI to complete assignments and that teachers can no longer give the same kinds of assignments. The district says families will be notified before any pilot tool is used and can request alternative assignments that meet the same learning goals.
Put those three stories side by side. Students are using AI every day. Teachers are redesigning on the fly. Systems are still arguing about the rules.
What AI assessment redesign reveals about our systems
In Chapter 1 of Human Still Required, I argue that AI didn’t break school. It exposed it. This week’s stories are a clean example. Take-home essays and homework packets were always proxies for learning. They worked well enough when producing the work required doing the thinking. That link is gone, and teachers are the first people who noticed.
Chapter 6 is where the teachers in the Reuters story are already living. Output is cheap now. What still counts is evidence of thinking, and I use the D.E.C.I.D.E.R. framework to name it: Decision-making, Explanation, Critique, Inference, Development, Evaluation and Relocation, meaning transfer to new contexts. Look at what these teachers landed on. An oral exam is Explanation. Asking which method a student prefers and why is Decision-making and Evaluation. A month spent on one essay, with drafting in class, is Development.
A student can generate an answer. They cannot fake understanding in a conversation.
Human Still Required
Many teachers worked this out on their own. That’s the good news.
Here’s the harder part. Some of the responses are designed to outsmart the tool: prompts AI handles poorly, sources AI can’t reach, laptops put away. I understand every one of those choices, and I don’t blame any teacher for making them. But in Chapter 2, I caution leaders against a retreat into tasks that are simply harder for a machine to complete. Tools improve. A paywall or a tricky prompt is a temporary barrier, not a definition of learning. Harder is not better. Clearer is better.
And that brings me to Chapter 12. When every teacher invents a separate answer, students get a patchwork. Rules vary from room to room. The definition of “your own work” changes between second and third period. Students notice, and they adapt to the inconsistency, not the learning.
The goal of professional learning is no longer consistency of practice. It’s consistency of thinking.
Human Still Required
I’m not asking leaders to make every classroom look the same. Chang’s oral exams and Wong’s in-class practice don’t need to be identical. They need to share a common answer to one question: what counts as evidence that a student did the thinking? That answer should be built together, not discovered alone.
What leaders can do now
- Find the redesign that’s already happening. Before writing a single new policy, ask your teachers what they’ve changed this year and why. The innovation is probably already in your building. It’s just invisible.
- Name a shared evidence standard. Use D.E.C.I.D.E.R. or your own language, but agree on what counts as evidence of learning across departments. Then let teachers choose how to gather it.
- Build time for peer redesign. Give teams protected time to bring one assignment, test it against that standard, and redesign it together. One redesigned task per team per month will do more than another tool training.
- Audit homework honestly. If a take-home task can be completed by AI in seconds, ask what decision it informs. Several teachers in the Reuters story have already moved practice into class. Make that a school conversation, not a private workaround.
- Talk to students about how they use AI. The NPR story shows students using chatbots as study partners and for feedback, not only to shortcut work. Ask them what helps their thinking and what replaces it. Their answers belong in your redesign.
If your leadership team wants to work through these questions together, the Human Still Required 5-Week Book Study is built for exactly this kind of shared thinking.
Frequently asked questions
How should schools redesign assessments for AI?
Start by deciding what counts as evidence of learning, then design tasks that surface it. Oral explanations, in-class drafting, conferences and transfer tasks show what a student can explain, defend and apply. The goal isn’t to block AI. It’s to make the student’s thinking visible.
Should teachers stop assigning homework because of AI?
Not necessarily. But any take-home task that AI can finish in seconds no longer tells you much about learning. Many teachers are moving practice into class and using homework for preparation, reading or reflection. Leaders should help teams decide together what homework is for.
Is it a problem if every teacher handles AI differently?
Different practices are fine. Different definitions of learning are not. When rules and expectations change from room to room, students learn to navigate the inconsistency instead of the content. Aim for consistency of thinking across classrooms, not identical lessons.
Closing
The teachers in this week’s news are doing what good educators always do. They saw the old proxy stop working and found a better way to see student thinking.
Our job as leaders is to make sure they don’t have to do it alone. Redesign shouldn’t depend on which teacher a student happens to get.
So here’s my question for you: if I walked into five classrooms in your school tomorrow, would I hear five different answers to what counts as learning, or one shared one?
Sources
- How teachers are remaking classrooms for the AI age — Reuters (Helen Coster), October 28, 2026
- Nearly 2 out of 3 teens and tweens use AI chatbots, new research shows — NPR (Rhitu Chatterjee), October 30, 2026
- Advocates call for AI moratorium to extend to NYC high schools — NY1 (Emma Barnett), October 28, 2026