The state of AI innovation in America's schools
Insights from the inaugural year of the AI Innovation Index (SY25-26), a national measure of how PreK-12 school systems are cultivating AI empowerment and agency with staff and students.
Skip to recommendations for funders and policymakersLeadership is the lever.
Student agency is highest where staff feel empowered. Staff feel empowered where leadership has executed key actions.
Effective leadership
District AI leadership: vision, policy, culture, and practice.
Empowered staff
Staff experience support and efficacy with AI in their roles.
Student agency
Students show judgment and agency in using and building with AI.
Explore the
SY25-26 findings.
Leaders are engaging stakeholders and progressing on executive-controlled actions
Districts have set direction and started listening. Changing culture and daily practice is where the field is stuck, a pattern that holds in 80 of 96 participating systems.
Direction-setting and listening run ahead of culture and practice
Mean rubric score by leadership domain, 1–4 scale. The split is not close: both direction-setting domains outscore all three culture-and-practice domains, and the ordering holds in 80 of 96 systems.
Actions a central office can finish with a decision or a document have been the first to move.
All 18 leadership actions, ranked by mean rubric score, 1–4 scale. The lagging actions require deeper capacity-building, cultural change, and shifts in classroom practice.
Top quartile leadership systems are furthest ahead in vision, engagement, and instructional workflows.
The single widest gap is Amplify Power-Users (+1.52 on a 4-point rubric): leading systems find their early adopters and put them to work, while everyone else leaves that energy on the table.
Positive outliers exist, even for leadership actions with the least national progress.
Even Hire for AI Competency, the lowest-ranked action nationally, has 11% of systems at active or advanced, so none of this work is waiting on conditions that don’t exist yet.
Leadership strongly predicts staff AI empowerment
What 7,144 staff across 45 systems report about working with AI, and how closely it tracks the leadership work above them. On personal questions, staff look similar everywhere. On whether their district is leading on AI, systems are worlds apart.
Where leadership scores higher, staff report more empowerment
Each dot is one school system; r = .67, p < .001. This is the strongest relationship anywhere in the Index, stronger than any single action or domain on its own.
Systems differ most on whether the district encourages AI innovation
Range of system-level scores by staff question, lowest to highest system (● mean). Personal questions cluster; the district-facing question splits the field, which is what you’d expect if leadership is the differentiator.
Staff in the best-led systems lead on every dimension
The widest gap (+31 points) sits on the one question leaders most directly control: whether the district encourages responsible AI innovation.
Central-office staff report more empowerment on every measure
The gap is smallest on retention (+9.6) and largest on team discussion (+23.4), which suggests school staff aren’t more skeptical of AI so much as less exposed to it in daily work.
Fall to spring: enthusiasm dipped, confidence in systems did not
Change in % agree from fall to spring. What softened is how staff feel about AI personally, not what their systems put in place around it. The likeliest story: novelty wearing off and a cooler public mood on AI.
Staff AI empowerment is the strongest signal of student AI agency
What 24,564 students across 32 systems report about understanding, using, and building with AI. Students broadly understand that AI matters. Far fewer can use it well, and fewer still can build with it.
Students understand AI far more than they can build with it
% of students who agree, by capability. Nearly twice as many students understand AI’s impact as can design and build with it, and the drop is steepest at the step that requires instruction rather than exposure.
Age doesn't build the skill. Teaching does.
High schoolers barely out-report middle schoolers, and on designing and building with AI, not at all.
Where staff feel empowered, students build
Each dot is one system. Staff are the adults closest to students’ daily work with AI, so it makes sense that their empowerment is the more proximate signal.
The gap is widest exactly where it matters most
Students in top-quartile staff-empowerment systems lead on every dimension, widest on designing and building with AI (+10.7 points). Awareness that AI matters, already near universal, barely moves.
Most school AI use stays basic, but innovative use is emerging, especially in central offices
What 2,900 staff free-text responses reveal about how AI actually shows up in the work.
Staff AI use concentrates in planning, communication, and assessment
Lesson planning alone drew 966 mentions and appears in nearly every participating organization, a use case that has barely changed since the first classroom AI tools arrived.
- Lesson and unit planning966 mentions · 42 of 46 orgs
- Parent and family communication707 · 42 orgs
- Assessment and quiz creation434 · 31 orgs
- Differentiation, IEP, and multilingual support314 · 23 orgs
Innovative use is emerging but concentrated in a minority of systems
The most advanced theme, building tools with code, appears in fewer than half of participating organizations, and the rest of these themes are rarer still.
- Building tools and automating with code109 mentions · 21 orgs
- Building custom AI assistants and GPTs76 · 10 orgs
- AI analysis of intervention and student-services data22 · 11 orgs
- Deep document analysis and synthesis11 · 7 orgs
Mostly in central offices.
Central-office staff use AI across a wider range of tasks than school staff
% of each group's responses mentioning the theme. Wider role-relevant use tracks with central-office staff's stronger AI sentiment, and leaves direct experience with the most advanced AI capability furthest from students.
Sentiment tracks what AI is used for
Most-mentioned use theme by school sentiment quartile. High-sentiment schools lean instructional; low-sentiment schools are far likelier to flag student cheating.
Capacity-building erases the AI leadership poverty gap.
Nationally, low-poverty school systems showed more AI leadership progress than their higher-poverty peers. But higher poverty systems participating in AI for Equity's AI Leadership Accelerator outperformed their lower-poverty peers nationally.
Partner systems opt into the program, so this comparison is descriptive, not causal.
High-poverty partner systems (2.17) out-score even low-poverty non-partner systems (1.70). District leadership rubric, 1–4 scale.
The most-adopted actions are not the ones most tied to better outcomes
The work a C-suite can finish with a decision or a document is getting done first. The work that requires change in culture and classroom practice is what distinguishes the leaders.
Done first. Separate systems the least.
Most tied to staff and student outcomes.
Evidence-based recommendations
What funders and policymakers can do to cultivate AI empowerment and agency amongst PreK-12 staff and students.
Leader capacity compounds. Fund it year over year.
Student AI agency runs through staff, and staff empowerment runs through leadership. Systems above the median on staff empowerment see student agency 14 points higher; systems above the median on leadership see staff empowerment 17 points higher, and ten of the eleven top-quartile staff systems have above-median leadership (leadership to staff r = .67, staff to student agency r = .60). That leadership capacity builds with sustained investment: systems with no Accelerator history average 1.75 on the 4-point rubric, first-year participants 1.92, and multi-year participants 2.45.
Shift assessment and accountability to unlock graduate profile work
Evolving the Portrait of a Graduate ranks near last of 18 leadership actions (1.57), and 58% of systems have not begun it, even as compliance-legible actions like task forces (2.54) lead adoption. A reasonable hypothesis: systems orient their work around the traditional outcomes their assessment and accountability systems reward. If applied AI literacy is going to reach graduate profiles and student programming, those incentives have to shift. The stakes are highest for systems serving students furthest from opportunity, often the most beholden to existing accountability structures.
Incentivize student-facing, applied AI literacy
Student AI literacy thins out as it becomes applied: 78.5% of students understand AI's benefits and harms, 58.5% understand how AI systems work, 54.6% have learned strategies for effective use, and only 44.8% have learned to design and build with AI (n = 7,220). The applied end does not arrive with age (high school 44.5%, middle school 45.2%). It arrives where adults are equipped to teach it: systems with above-median staff AI efficacy see students 11 points higher on design and build.
Fund the field's measurement infrastructure
Every finding on this page exists because leadership, staff, and student measures were collected in the same systems, via the same infrastructure, at the same time. That is what system-level measurement infrastructure like the AI Innovation Index makes possible. This allows linkages from the adults doing the work to the students that work is meant to reach, allowing us to move from AI anecdote and assumption to AI evidence.
Sign up to participate.
The SY26-27 cycle is open and participation is free.