Artificial Intelligence

Across our investment areas, NewSchools funds early-stage ventures that use AI to catalyze better teaching and learning. We believe AI can be a powerful tool when grounded in research, co-created with educators, and focused on what works in real classrooms.

2025 Investments

In 2025, we expanded our AI portfolio by supporting 22 new ventures, bringing the total number of AI-powered solutions to 47 since 2019. Outside of our schools portfolio, 70% of funded ventures leverage AI to improve outcomes for students furthest from opportunity. This year, we also launched the GenAI Math Tutoring Program, which provides research & development support for developers to keep quality, evidence, and inclusion at the center of their design.

$4.9M

invested

22

funded ventures

53%

led by former educators

Impact Highlight

AI-enabled solutions with rigorous evidence reported increasing K-8 math and English Language Arts (ELA) student learning by 13% on average.

What We’re Learning

AI works best when designed for people, not efficiency

AI can reduce administrative burdens, strengthen instructional feedback loops, and help educators focus on what matters most: inspiring curiosity, building relationships, and helping every student believe in their potential.
Early-stage pilots can drive evidence-based design

As interest in AI grows, there is an opportunity to rethink how early-stage education research drives impact and growth. By combining user-centered design and classroom pilots with tools like AI benchmarking, ventures can generate decision-relevant evidence earlier, more safely, and at lower cost.
AI can support deeper thinking and learning

GenAI math tutoring is shifting from a focus on procedural correctness toward supporting deeper, more holistic student learning. Ventures that attend to how students think and learn, not just the steps they take, are well positioned to redefine effective math support and drive more equitable impact at scale.

Venture Spotlight

What if the Best Way To Learn Math Is To Teach It?

For many students, math becomes something to memorize and move past. Research suggests deeper learning happens when students explain ideas and teach others. ALTER-Math was built on that insight.

Developed by researchers at the University of Utah in collaboration with multiple institutions in the U.S., ALTER-Math invites students to teach a simple AI agent what they know. As students explain concepts and correct misunderstandings, the agent updates its knowledge in real time. This process encourages reflection, conceptual understanding, and active learning.

Led by Dr. Chenglu Li, whose own learning was shaped by teaching peers, the team has aligned ALTER-Math with Math Nation, a widely used, standards-aligned curriculum. In early pilots with tens of thousands of students, those using the teachable agent showed nearly 50% greater learning gains than peers using traditional tools.

ALTER-Math is designed with access in mind. The tool is offered at no cost, with plans to open-source the technology so others can build on it. As states explore how to use AI responsibly in classrooms, ALTER-Math offers a model that supports learning by putting students in the role of active thinkers and teachers.