In February 2026, Daniela Amodei, President of Anthropic, sat for a Fortune interview and said something her industry doesn’t usually say out loud. The skills that matter most in the AI era, she argued, are the ones AI doesn’t have. The humanities. Critical thinking. The ability to interact with other people. She called them “more important in the future rather than less” (Daniela Amodei, Fortune, February 2026).
The president of one of the most influential AI companies in the world told the business press that human skills are the curriculum. That’s worth pausing on. Most school-level conversations about AI are still organized around the wrong axis: whether AI replaces teaching, whether to ban it, which subjects it threatens. The sharper question is the one Amodei pointed at: what kind of teaching has the system actually been doing? And what kind has it been quietly leaving out?
What AI Made Impossible to Keep Ignoring
For decades, education has measured itself on a narrow slice of cognitive work. Rote memorization. Multiple-choice grading. Standardized tests scored at factory scale. There was an operational reason for it. Multiple choice grades fast: A, B, C, or D goes in and a number comes out. As longtime collaborator and early-stage investor Danh Trang put it, “it’s like factory: in, out.” You can grade a hundred bubble sheets in an afternoon. You can’t grade a hundred speeches in an afternoon.
So the curriculum bent toward what was countable. Recall over reasoning. Selection over creation. Compliance over critique. That bent worked, sort of, in a world where the bottleneck on knowledge was access. It works far less well in a world where AI generates fluent output at near-zero marginal cost.
What AI has actually exposed is that the cognitive skills the system was training were the lower-order ones. As Dr. Ilana Nankin, founder of Breathe for Change, has framed it, “there has been such an overemphasis in education historically on cognitive intelligence, rote memorization, standardized testing, and those are skills that are lower than critical thinking, decision-making, creativity when it comes to cognitive intelligence.”
The distinction is load-bearing. Students don’t need less cognitive development in the AI era, they need more, and of a different kind. Reading’s the clearest example. AI can read for you in seconds, and the barrier to producing content has collapsed at the same time, so the volume of writing landing in front of every student has gone up. Knowing how to read carefully, spot the implicit assumption, and decide what’s true for yourself becomes more urgent, not less. The skill the system under-trained is the skill the era now demands.
A 2025 systematic review of 31 studies on AI and higher-order thinking in education reached a similar conclusion. Positive learning experiences outweighed negatives when AI was used intentionally, but reviewers flagged that AI struggles precisely where higher-order thinking is hardest: evaluating depth, handling complexity, and developing genuine ethical reasoning (Patrick et al., Higher Education Quarterly, 2025). Higher-order cognitive skills don’t develop because AI is in the room. They develop because educators design for them.
Four Dimensions the System Has Never Really Taught
Fixing the cognitive curriculum is necessary. It isn’t, on its own, sufficient. Even if every classroom traded standardized recall for critical thinking tomorrow, the system would still be leaving four dimensions of human development almost entirely untouched.
Breathe for Change calls these the other four layers of the Human Intelligence framework: emotional, somatic, relational, and universal. They aren’t soft skills. They’re the dimensions of being a person.
Emotional intelligence is how students understand and regulate what they feel. The capacity to name an emotion is the first step toward not being run by it. A meta-analysis of 213 school-based social-emotional learning programs involving more than 270,000 K-12 students found that participants gained 11 percentile points in academic achievement alongside reductions in conduct problems and emotional distress (Durlak et al., Child Development, 2011). Emotional development reinforces academic achievement rather than competing with it.
Somatic intelligence is the intelligence of the body. Attention, breath, regulation, the felt sense of being present. It’s the layer that gets crushed first when classrooms become screen-mediated and worksheet-driven. It’s also the layer that determines whether a student can focus long enough to do higher-order cognitive work at all.
Relational intelligence is how students relate to each other. Empathy, repair, belonging, the capacity to be in conflict without breaking the connection. “Who do you want to hire in the future,” Dr. Nankin asked, “someone who has human skills in addition to the integration of technology in their lives?” Founders building the next generation of companies are answering that question in real time, and the answer is unambiguous.
Universal intelligence is connection to purpose, the reason a student does anything at all. “Education is in the moments between,” as Dr. Nankin put it, “an embodied experience of living, connecting, creating.” Take that out, and the rest is just compliance.
For a deeper look at one of these layers specifically, this piece on why emotional intelligence defines the next era of education, and what AI can’t feel builds the argument further. And what neuroscience reveals about the human capacities AI can’t replicate provides the evidence layer.
What This Looks Like Inside a Classroom
There’s a version of this argument that lands as abstraction. Five layers, five capacities, five lovely words. It has to land as practice instead, because the adoption data is moving fast. By spring 2025, more than 80% of K-12 students said their teachers hadn’t taught them how to use AI for schoolwork, even though student use had already crossed 50% (RAND Corporation, 2025). By December 2025, 62% of students in grades 6 through college were using AI for homework, a fourteen-point jump in seven months (RAND Corporation, 2026). Adoption isn’t waiting for the curriculum to catch up.
What does catching up look like? It looks like teachers modeling explicitly when AI belongs in a task and when it doesn’t. Classrooms where students do speeches more often, because AI can give them feedback between teacher conferences, and where they get more reps on the higher-order work. As Danh Trang observed, AI “allows the student to get so many more reps in” on the skills that used to bottleneck at one teacher, thirty students.
It also looks like classrooms where the four neglected layers get explicit instructional time. Emotional regulation taught alongside reading. Somatic practices threaded through the school day. Relational repair built into community agreements. Purpose surfaced and named, not assumed. The American Institutes for Research’s multi-year study of “deeper learning” network high schools, which centered higher-order cognitive skills and integrated student wellbeing into instruction, found that students graduated on time at significantly higher rates and were more likely to enroll in four-year colleges than matched peers (American Institutes for Research, 2014). The whole-person approach produced the achievement.
A composite voice from the field captures the shift well. “I’ve been told for years that test scores are the goal. Now I’m being told the test scores are what the AI does. So what’s left for me to teach?” The answer is everything that actually matters.
The Curriculum Architecture for the AI Era
The phrase Dr. Nankin uses for the relationship between teacher and AI is “thought partner.” It’s a precise phrase. The educator leads. The AI assists. The thinking belongs to the human. That relationship is what the five layers of the Human Intelligence framework are designed to make possible.
This is the curriculum architecture behind the William Jewell College Master of Science in Education (M.S.Ed.) in Transformative Teaching & Learning, in partnership with Breathe for Change. Higher-order cognitive development. Emotional, somatic, relational, and universal intelligence taught as instructional skill, not as an optional add-on. The same framework educators experience for themselves, they bring back to their classrooms. The whole-person approach isn’t theoretical. It works at the level of the building, the day, the kid.
A larger argument for reclaiming human agency as educators in the age of AI underpins this one. The further move is the cognitive nuance: the curriculum reckoning isn’t about choosing human development over cognitive development. It’s about doing both, fully, for the first time.
What Amodei Was Pointing At
The president of Anthropic told Fortune that human skills are the most important skills to cultivate in education going forward. She put that responsibility on educators. And she named skills the system has under-developed for decades.
The path forward isn’t to teach less in the cognitive lane. It’s to teach the higher-order parts of that lane the system was never set up to grade, and to finally treat the four other layers (emotional, somatic, relational, universal) as the rest of the curriculum rather than the extras.
This is the work.
The M.S.Ed. is built for educators who want to do it. HLC-accredited through William Jewell College, FAFSA-eligible, with estimated tuition of $14,790 before financial aid. Learn more about the M.S.Ed. here.
Want to go deeper? Listen to the full conversation between Dr. Ilana Nankin and her longtime collaborator and early-stage investor Danh Trang on A Work of Heart, the Breathe for Change podcast. They cover what AI exposes about the curriculum, the cognitive distinction between low-order and high-order skills, and the four dimensions of human intelligence the system has been under-teaching all along.
