The Market

The K-12 math EdTech market is attracting significant philanthropic and venture capital, but on fundamentally different terms. Philanthropic funders including Arnold Ventures, the Gates Foundation, and the Chan Zuckerberg Initiative have backed evidence-based adaptive platforms with independent evaluation requirements and near-term accountability for student outcomes. Venture capital is flowing toward autonomous AI tutoring models on the premise that generative AI will eventually produce outcomes that close the current evidence gap, with a longer time horizon and higher risk tolerance than most philanthropic funding mandates allow. That divergence in investor logic is the structural context this opportunity audit addresses. The core decision is whether ASSISTments' independently replicated, relatively modest learning gains constitute a stronger investment case for philanthropic capital than higher-profile AI-forward platforms whose market adoption currently outpaces their evidence base.

A 2026 PRISMA-guided systematic review of AI-supported teaching in secondary mathematics education, synthesizing 31 studies published between 2022 and 2025, found that AI's primary significance lies more in redistributing teacher work than automating it. Stanford HAI's analysis of AI scaffolding in math classrooms illustrates the mechanism, showing that AI helps teachers build and personalize scaffolds that were previously too time-consuming to produce at scale, without removing teacher judgment from the workflow. A case study published in Frontiers in Education examining undergraduate mathematics instruction found that students relying solely on ChatGPT performed significantly worse than those receiving teacher-led instruction or teacher instruction augmented with AI. A Purdue scoping review on AI in K-12 education identifies teacher agency, ongoing professional development, and systemic support as the factors that determine whether AI tools produce outcomes worth funding.

These four sources do not prove that autonomous AI tutoring always fails. They establish that the current evidence most consistently supports teacher-centered augmentation as the model with the strongest outcome record, which makes it the appropriate baseline against which to evaluate alternatives. 

That said, this analysis represents a current conditional snapshot, as the autonomous AI tutoring tools compared here are recent enough that longitudinal outcome studies remain limited or absent, and the underlying technology is advancing rapidly enough that today's findings may not fully reflect the systems funders will encounter in the near future. A reassessment would be warranted as both the evidence base and the technology itself mature.

Screening Exhibit: The Decision-Relevant Math Learning Opportunity Set

This exhibit narrows from the full market to the four models most relevant to the investment decision. 

The four platforms selected here represent the decision-relevant range of solution architectures a philanthropic funder is actually choosing between, from AI-enabled autonomous delivery to teacher-mediated delivery. ASSISTments is a free teacher-integrated math homework and feedback platform. ALEKS is a subscription-based adaptive math practice system used in both supplemental and standalone modes. Khanmigo is Khan Academy’s AI tutoring assistant, deployed as an optional or integrated classroom tool. Alpha School is an autonomous AI-delivery private school network replacing traditional teachers with guides. 

Platforms including Brilliant, Photomath, and Synthesis each signal adjacent market directions (premium interactive learning, solver-first assistance, and AI-native tutoring, respectively) but are less central to the investment decision addressed here.

Five Investment Criteria

Evidence Quality

Does independent, peer-reviewed evidence of learning gains exist, and has it been replicated?

Implementation Conditions 

Do gains depend on specific deployment conditions, and are they realistic at scale?

Equity Relevance

Do demonstrated gains extend to diverse and under-resourced student populations?

Cost-to-Gain Ratio

What does a funder actually pay for, and what is the practical magnitude of the return?

Scalability

Can the model expand without degrading the conditions that produced its evidence?

Each criterion is applied to ASSISTments and three decision-relevant platforms in the analysis.