Four Priorities (not Predictions) for 2026

In June of 2024, I identified four priorities for human-centered AI in schools: augmentation over automation, literacy over policy, design over technology, and vision over decisions.

These priorities have become the heart of my work. They shape how I write, curate content, design workshops, and advise schools on how to approach AI. Every few months, I poke at them, wondering if they’re still relevant, if their meaning has changed.

The beginning of a new year is a good time for this reflection, and I thought I’d share how I’m thinking about my own AI priorities in the coming year. I’m evolving on a lot of these points, but one thing has not changed: as AI improves, the need to prioritize human values and behavior in our exploration of AI only becomes more important.

When I first wrote about this priority, we were deep in the “human in the loop” phase of generative AI, when any effective use relied on the human’s ability to prompt, evaluate, and coach the bot. Augmentation was a requirement for effective use of large language models; without us, LLMs were likely to produce unusable content. This requirement had an added benefit: we learned how to work with bots as tools, how to navigate their weaknesses and take advantage of their strengths.

This was before agentic AI capabilities became widely available and before LLMs could do more accurate and high-quality work more quickly. This year, I think we should anticipate more delegation of tasks to AI with minimal human intervention. In other words, more automation.

I recently wrote about agentic capabilities in AI tools and included several demos of how we can now assign tasks to LLMs. Recently, I went to NotebookLM and asked it to create a slide deck based on my most recent article, “Nine AI Messages for Students.” I did nothing other than provide NotebookLM with a link to the piece and click the “Slide Deck” button in the Studio section.

In about five minutes, it produced this deck. Not only is it accurate, it is a useful distillation of the article’s points, and NotebookLM created its own metaphors to explain those ideas (blueprint and compass… not creative, but certainly not incorrect). It leaves nothing major out.

My experience with AI this past year has led me to see augmentation as less of a technical requirement and more of a personal choice. How do we want to use the power of AI to serve our human goals, not just automate a task we would rather not do? How do we align our use to our values? If AI saves us time, lightens our cognitive load, or sparks new ideas, what do we do with that added capacity? Augmentation should go beyond efficiency and prioritize effectiveness, ethics, and innovation.

At the heart of this is a “human first, human last” approach. Take the deck NotebookLM created for me as an example. I spent hours writing the article it is based on; in this case, I basically used it as detailed, organized prompt for AI. I didn’t just provide NotebookLM with an idea; the article provided it with examples, argument, and organization. And, because I wrote the article, I can use my judgment to assess the output. All NotebookLM did was save me about a half a day of compressing paragraphs into punchy headlines and bulleted lists, choosing images, and organizing slides. It is an extension of my own work.

Slide decks are pervasive in schools. Teachers use them to deliver instruction, students are asked to make them for assessments, administrators use them to communicate ideas, etc. A well-composed deck is a product that signals many things about a person: effort, critical thinking, creativity, care, design sense, personal taste, etc.

As it has done with many forms of writing, AI has weakened the slide deck’s reliability as a signal of human effort. What are other signals, besides a polished deck, that we can look for to assess human thinking? What about our process needs to be more visible and explicit now that AI can easily create products? What kind of work matters now? The answers to these questions will define how we differentiate between augmentation and automation.

In the spirit of human-centered, relational school cultures, I would love to reframe AI guidelines as acts of care rather than acts of control. That shift requires more transparency and mutual understanding. A first step for both teachers and students is to become more literate about each other’s experiences.

For example, here are some notes from time I’ve spent with students over the last few months:

• Students are aware of which of their teachers use AI detectors, and they know how to hide their use of AI from those detectors. Students who don’t use AI run their original work through AI detectors to see if they might be accused of cheating when they have not (and sometimes, sadly, they make edits based on the detector’s assessment). Furthermore, students are running their teachers’ work through AI detectors, things like feedback, narrative reports, and class materials, to see if teachers are following their own rules.

• Students from AI-literate families have impressive knowledge about AI, talk about it with their parents, and use it in a wide variety of ways because they are surrounded by people using it. Students from AI-literate families who choose to resist the technology can speak fluently about ethical concerns. Students who are not exposed to AI in the same way tend to rely on what they learn from peers, the internet, and their own experiments.

• The most common use students share with me is AI as study buddy. They use it to create flashcards, make sample quizzes and tests, ask them questions, break down concepts, etc. When I ask students what job AI does for them, “tutor” and “assistant” are the words that come up the most.

• Students from well-resourced families have premium subscriptions to chatbots (or they use their parents’ accounts), giving them access to sophisticated models and new features that other students do not have.

• Students’ use of AI extends beyond school. They often use it as a toy or as entertainment, having it write or tell them stories or create images or videos that make them and their friends laugh. They use it as they would Google, for information on topics that interest them. They use it for hobbies, like coding or design. They use it to apply to internships or to build resumes and social media profiles. They know to look for errors, hallucinations, or “robotic” quirks in the output.

• When students start exploring AI for companionship (advice, therapy, comfort, etc.), they often tell me they do it out of desire for privacy and for convenience.

• Students are frustrated by their schools’ focus on cheating when they have bigger questions about AI, like how it will affect their career prospects, what it’s doing to the environment, where its most exciting uses are, and where its scariest uses are.

I wonder how aware teachers are of these experiences, and I wonder if the discourse on AI in schools makes room for all of these experiences. We would benefit from looking at AI through students’ eyes to better understand the many ways they interact with it.

For teachers, their primary concern is that they have an authentic understanding of their students and their students’ work, and they see AI use as an obstacle to that understanding. While students understand their work needs to be “good,” they don’t often see their assignments as opportunities for authenticity. Many students I meet conflate quality with polish: something that looks or sounds “professional” would get a better grade than something in their own voice, which might seem “imperfect” or “amateur.” I think students would benefit from seeing more examples from their teachers of what learning looks like, as opposed to performance.

Much of this boils down to more conversations about why we do the work we do in the classroom, and what AI does and does not have to do with that work. Unfortunately, many of those discussions don’t happen until after a student has broken a rule. Yet, as Torrey Trust and Robert Maloy explain, classroom policies that proactively include why certain rules are in place are not just more motivating to students, they are also ways for teachers to communicate purpose and share their values.

The most common AI policy I see in schools is some version of “you may use AI only with the permission of your teacher.”

A few assumptions underlie this policy:

• The teacher understands what the school means by “AI.”

• The teacher is able to detect both prohibited and permitted AI use.

• It is only the student’s integrity that is in question, not the design of the assessment.

• AI is the only form of external assistance that should be regulated.

I have written before about shifting the essential question about AI and assessment from “How do we stop students from using AI?” to “How do we know students have learned?” Decisions related to assessments should prioritize the second question, not the first.

I’ve started to think about Leon Furze’s “Five Principles for Rethinking Assessment with GenAI” as a roadmap. Like Furze, I believe schools should respect teachers’ professional judgment. I also believe schools should work with teachers to ensure that judgment is built upon a deep, shared understanding of validity, authenticity, transparency, and the purpose of assessment in and beyond the context of AI.

Focusing only on student behavior and deploying imperfect technological interventions do not confront the core issue: the emergence of AI has created both new issues with certain assessment models and exacerbated old ones. Whether a teacher decides to explicitly integrate AI into an assessment or to use analog, supervised methods to resist it, I believe the first four principles should be prerequisites for the fifth.

In my experience, teachers have far more expertise in and excitement about the topic of assessment than they do AI. AI may not be the catalyst for change that we would have wanted, but it presents an opportunity to take a serious look at our assessment models, evaluate their validity given the arrival of AI, and work together to make adaptations that serve student learning.

In their 1994 essay “The Grammar of Schooling: Why Has it Been so Hard to Change?” David Tyack and William Tobin explore the “grammar” of school, which they define as “the regular structure and rules that organize the work of instruction” like age-based cohorts and grades as evaluation tools and the organization of the academic program into siloed disciplines. This grammar is so entrenched that we take it for granted, making it challenging to interrogate, much less change.

But, it has changed over time, if slowly, and as we strategize for the uncertainty and rapid change that define AI right now, I think Tyack and Tobin’s observation about how change has actually happened in schools is important (emphasis is mine):

“Reformers believe that their innovations will change schools, but it is important to recognize that schools change reforms. Over and over again teachers have selectively implemented and altered reforms. Rather than regarding such mutations as a problem to be avoided, one might entertain the notion that they are potentially a virtue—reforms might be designed to be hybridized according to local needs and knowledge. Likewise, goals themselves might be regarded as hypotheses—pragmatic blueprints to be evaluated by their effects—rather than as fixed targets.”

When it comes to AI, so much is outside of our locus of control. But, schools can control what they communicate to their employees and students about how to approach AI. Empowering people to recognize the challenge AI presents, to articulate and test hypotheses about how to respond, and then to collectively evaluate the effects of those experiments and develop strategies that suit the school’s unique identity is a way to assert our agency, to focus on agility rather than control. This is action research. This is improvement science. This is reflective practice. This is authentic assessment for adults. Whatever you want to call it, it’s a strategy built on human expertise, creativity, and collaboration.

The first step in adopting this strategy is acknowledging that adaptation in the face of AI is required of us, and we must become comfortable in a fluid environment where revisiting and revising decisions is proactive, not reactive. What if we can’t resolve the big questions right now? What if we can only explore, learn, and adjust accordingly?

• If you want to learn more about my work with schools and nonprofits, take a look at my website and reach out for a conversation. I’d love to hear about what you’re working on.

• February 25. Kawai Lai and I will be facilitating a three-hour workshop, “Human First, AI Ready” at the NAIS annual conference in Seattle, WA, USA. This workshop is designed for school leaders who are navigating the complexities of AI integration at school, including defining ethical behavior, navigating diverse perspectives, and supporting a strategic and sustainable approach.

• June 16-18. I’ll be facilitating a three-day AI program called “Learning and Leading in the Age of AI.” This intensive residential program is designed for school teams to have time and space to design classroom-based and school-wide AI applications for the next school year. Hosted in partnership with the California Teacher Development Collaborative (CATDC) at the Midland School in Los Olivos, CA, USA.

• June 23-26. I’ll be joining the Summer AI Institute at Lakefield College School (Lakefield, Ontario, Canada) as a speaker and coach. This event is for teams of educators from Canadian independent schools to advance their AI work, design classroom and school-wide AI initiatives, and learn from each other’s work.

• January 7 to April 1. I’m facilitating “Leading in an AI World,” a four-part online series for school leaders navigating the complexities of AI integration at school. We’ll review the AI landscape, look at how to shape mission-aligned position and policies, and explore a variety of ways to engage colleagues, students, and families in meaningful AI learning experiences. Offered in partnership with the Northwest Association of Independent Schools.

• January 22. I’ll be doing a second run of “AI and the Teaching of Writing: Design Sprint” in partnership with CATDC. This workshop, designed for teachers of writing in all disciplines, offers some important considerations, practical examples, and hands-on exercises to consider how we should adapt the teaching, practice, and assessment of writing in an AI age.

• Given what happened in Venezuela this week, it’s important to remember that high on the list of scary applications of AI is its use in warfare and other military operations.

• Two of the most interesting essays I read in 2025 were “AI 2027” and “AI as Normal Technology,” which present diverging visions of the future of AI. Now, the authors have come together and written this piece about their points of agreement. Helpful for those trying to find signals in the noise of AI predictions.

• This preview of Google Disco offers a hint of how AI is going to change how people navigate the internet.

• In her new TED Talk, Sasha Luccioni explains her research on the environmental impact of data centers and offers a vision for a more sustainable future for AI: smaller models trained for more specific work.

• Fei-Fei Li, a pioneer in AI development, highlights ongoing work on AI’s spatial intelligence, an innovation that will take us beyond LLMs.

• New York University professor Panos Ipeirotis brought oral exams into his class to assess students. The twist? He developed an AI voice agent to conduct the assessments and used AI to support grading the assessments. It didn’t go perfectly, and he wrote a compelling, warts-and-all reflection on this experiment.

• I found “Large AI Models are Cultural and Social Technologies,” co-authored by a psychologist, a data scientist, a sociologist, and a political scientist, so helpful in thinking about how to engage AI in a human-centered way.