The most powerful innovations in education emerge when researchers, developers, and educators work side by side. That's why Leanlab Education is partnering with NewSchools Venture Fund on an ambitious research initiative exploring how generative AI (GenAI) can transform math tutoring for K-12 students.
The Partnership
Earlier this year, NewSchools launched a new early-stage GenAI math tutoring solution initiative and selected Leanlab Education as a research and evaluation partner. NewSchools reviewed over 260 proposals for early-stage GenAI math tutoring solutions and selected seven promising ventures to support: Thinkverse, Snorkl, Goblins, University of Utah’s ALTER-Math, University of Colorado Boulder’s Talk Trees, Step Up Tutoring, and CitySchools Collaborative. Leanlab, with the additional expertise of The Learning Agency, is working alongside these ventures to ensure that each tool is not just technologically sophisticated, but truly ready to make a difference in classrooms.
The timing of this work is critical. Math scores remain in decline, GenAI use is rapidly increasing in the classroom, and schools are looking for effective ways to reach more students. This initiative aims to harness the power of GenAI through R&D to personalize learning and support student success, while helping the field better understand the impact and potential of new technologies.
Coming Together: A Community of Practice
In mid-October, NewSchools convened their new cohort for a two-day Community of Practice, during which the seven GenAI math tutoring ventures gathered with the Leanlab team to learn from one another and prepare for the work ahead. With classroom pilot studies planned for the spring, the gathering laid the groundwork for strong implementation and the collection of high-quality data.
The first day featured a panel moderated by Leanlab's COO, Amber Daniels Hatz. The panel brought together different perspectives on what it takes to bring AI tools successfully into schools. Kristi Preston, Managing Director at Watershed Advisors shared insights on navigating the complex policy and funding landscapes of states. Dr. Carolyne Quintana, CEO of Teaching Matters and former Deputy Chancellor of Teaching and Learning for NYC Public Schools, spoke about the real-world challenges of implementation at scale. Matt Miller, co-founder and CEO of OKO Labs, offered the developer perspective on building tools that truly meet educator needs.
The conversation explored questions that every venture grapples with. How do you avoid the "5% problem," where only a small group of highly-engaged students benefit from a promising tool? How do you ensure your tool fits into the existing complex system of a classroom, and the teacher can confidently implement it?
On the second day of the Community of Practice, the Leanlab team led a deeper dive into codesign research and educator collaboration. The ventures spent time learning about the other tools, sharing challenges they face and solutions they have found. They also met with local teachers to hear directly about their experience with AI in the classroom, including their fears and desires for safeguards that can ensure the safety of GenAI tools.
Across conversations, a powerful theme emerged: the need to balance innovation with coherence. These ventures aren't just asking "Does our tool work?" They're asking "How does our tool fit into the larger picture of what teachers and students need?" They’re considering curriculum alignment, implementation fidelity, the right dosage, and interoperability with other systems.
Leanlab researchers and the ventures also kicked off their year-long codesign research process at the Community of Practice, meeting with the educators set to engage in usability and efficacy studies of each tool. At the end of the week, the ventures left ready to dive into their research journey, with a new community of peers and partners.
What’s Next
Research is already underway for the ventures. Leanlab researchers are working with the ventures to conduct research audits, develop logic models, and benchmark each AI tutor's mathematical accuracy, pedagogical quality, and fairness.
This fall, we’re conducting rapid cycle evaluations with educators from Leanlab’s AGILE Network. These iterative usability tests with educators provide the ventures with actionable feedback to refine their products. Everything we're doing this fall helps ensure that each tool is ready for classroom pilots launching in spring 2026.
Why all this preparation ahead of the spring efficacy research? If a tool freezes frequently, or if the user experience is confusing, it won’t be used with fidelity. Without fidelity, come spring we cannot collect the high-quality data needed to understand how each tool impacts student learning.
By summer 2026, this research partnership will provide a deeper understanding of best practices for building, adopting, and evaluating GenAI math tutoring solutions.
The future of AI in education will be shaped by partnerships like this one, where researchers, developers, educators, and investors work together to test ideas, learn from real classrooms, and build solutions that truly serve students. We can't wait to share what we learn along the way.