Editor’s note: This essay is an entry in Fordham’s 2026 Wonkathon, which asked contributors to answer this question: “What rules should govern the uses of digital technology in schools—and who should set them?” Learn more.
I’m an instructional nerd and a professional learning person. I was never a blended learning aficionado, but I have seen two things in the past few years that make me believe that AI-powered technology will be an important part of the future of teaching and learning:
1. I saw a product team improve the quality of feedback that their product gave students on writing assignments faster than the school could schedule a professional development day for teachers.
2. I saw a product, on its own, identify how to help more students better understand a frequently missed question.
I believe that AI-powered tools will be part of the future of teaching and learning partly because of the value that I already see them delivering but more so because of the rate of improvement that they have already shown compared to what we have historically seen in instructional improvement efforts. I have firm conviction that teachers will continue to be the central leaders of teaching and learning and that instruction will continue to be fundamentally a relational act. I also have firm conviction that, though we’ll have to manage the significant risks that AI creates in classrooms, AI-powered tools can make it easier for teachers to be great and for students to learn more.
Financial investors have an old rule: Don’t bet on the concept, bet on the team; above all, bet on the leader. School system leaders should bear this adage in mind as they make product purchase decisions. Given how quickly AI-powered tools are changing, when a district buys a tool, they’re choosing a partner as much as access to software. The way that partner operates will determine whether the product gets better or worse for students over the life of the contract.
I’ve been privileged to spend time as part of Instruction Partners’ AI in Action Learning Tour talking to product developers about what problems they’re trying to solve, how their tools work, and what they’re still figuring out. (You can watch interviews with product developers on their respective profile pages.) I have also been lucky to follow the development of several products over time and understand how leaders are guiding their evolution. Based on those experiences, here are five questions that I’d advise school system leaders ask product developers and what I would look for in strong responses:
1. How often is your senior product leader in classrooms and what happens after they visit?
The strongest product leaders that I’ve met are in multiple classrooms monthly, and they use these visits to drive immediate improvements. They are, frankly, obsessive. They examine every question, every tool response, and every shred of feedback—and expect their teams to make changes immediately. If I were interviewing a product leader, I’d be looking for them to name specific schools they’ve recently visited, something the team changed because of those visits, and probably a joke about how the product team has built an agent to simulate the leader’s obsessive attention to quality.
2. What is one popular prior feature in your product that you’ve sunset, and why did you stop it?
Students and teachers sometimes love features that don’t help students learn. For example, they want releveled texts at an instructional moment that requires scaffolds or they like a stop and jot that feels pretty easy but takes a lot of time. A team willing to remove a popular feature because student work showed it undercut learning or wasn’t worth the opportunity cost is a team that holds itself to outcomes rather than engagement. I’d be looking for developers to name a feature they cut, the evidence that prompted their decision, and what, if anything, replaced it. A team that has never cut anything, or cut features only for technical reasons, is likely not realistically evaluating the trade-offs involved in driving student learning.
3. How have you improved the feedback that your product gives students, and what are you still working to improve?
AI makes individualized feedback possible at scale. It also makes it easy to produce feedback that is inaccurate, too long for young readers, or that does the thinking for students. Improving the feedback a tool offers is forever work—there will always be something to make better. The strongest teams review samples of their feedback against defined standards, often with experienced teachers, and can name specific failures they found and fixed. I’d put the most credence in teams that are proud of what they have improved and honest about where they are still tuning. A team with no open problems in its feedback probably isn’t looking closely enough.
4. What have you learned about what high-use teachers with strong student outcomes do differently than high-use teachers whose students make less progress?
The same product will be used in a range of ways; even products with strong evidence of impact can’t guarantee that they’ll deliver those outcomes in every classroom. A team that can answer this question well cares about learning more than use and has studied its own implementation data alongside student learning gains. I’d be listening for understanding of which uses do not benefit students and what they did with those insights to steer users toward better use.
5. What customizations have you made for specific school systems, and what customizations do you refuse to make?
To work well for students, AI-powered instructional tools need to be coherent with the other learning tools the system already uses. For example, aligning practice problems to the system’s adopted curriculum and scope and sequence can help AI tools match the language teachers already use for feedback and ensure that activities align with the district’s assessment calendar. These kinds of adjustments are a sign of instructional strength. However, product developers who have parameters on what they won’t customize is also a sign of instructional strength. A team that has a clear theory of learning will better improve student outcomes over time than a team that is willing to adapt anything on request.
Across these questions, school system leaders should look for a product partner that has a point of view based on 1) experience, 2) research that demonstrates evidence of impact, 3) awareness that product design drives student learning only in part—implementation will matter just as much—and 4) healthy respect for the expertise that school system leaders and educators bring. The teams that I would bet on to drive substantive improvement are the teams that see every decision—from significantly reworking the feedback the tool provides to simply moving a button to the other side of the screen—as an opportunity to improve student learning.
