Spotlight on Innovation: Reflections on GenAI Investments and Trends in Education

Since ChatGPT burst onto the scene in 2022, there’s been a groundswell of attention and activity concerning generative artificial intelligence’s (genAI) potential to transform Pre-K-12 education. In this short period of time, many districts have taken, and in some instances already reversed, their positions on AI, as well as launched AI initiatives that have turned into cautionary tales. GenAI went from bad and potentially unable to improve at math by some estimations merely months ago, to now being capable of competing for gold in the International Math Olympiad. At flagship education and philanthropy conferences, it has gone from obscurity to inescapability.

As a funder supporting innovation and scalable solutions in education, Overdeck Family Foundation has intentionally explored the potential of genAI to advance our mission. Over the last two years, we’ve made over $13 million in commitments across 16 grants with the primary purpose of building, scaling, or learning about AI-powered interventions. All told, we’ve observed more than 20 grantees using AI in a variety of ways to deliver better, more cost-effective solutions that support students, teachers, and families.

Recently, our team took stock of our grantmaking efforts and trends we’ve seen in the broader landscape to inform how we pursue genAI investments in 2025 and beyond. Below, we share five reflections that surfaced focused on the progress, challenges, and dynamics we’re observing through the efforts of our grantees and others.

1. GenAI use is currently concentrated in adult-facing versus student-facing use cases

According to a recent Walton Family Foundation survey, 79 percent of teachers are familiar with ChatGPT—a sharp increase from 25 percent only one year ago. A separate RAND study found that nearly one in five teachers were using AI for teaching in fall 2023, but by the end of SY 2023-24, 60 percent of districts planned to train teachers to use this technology.

When compared to student-facing applications, we’re seeing greater adoption of AI-powered tools supporting adults—from teachers and tutors to early childcare educators—to perform their work more efficiently and effectively. While this may partly reflect greater data privacy and security concerns regarding student-facing use cases, the benefits for adults are already surfacing in two key areas. First, these tools save time by streamlining day-to-day tasks such as lesson planning, allowing teachers more time for deeper learning and relationship building.

Second, the emergence of genAI-based coaching tools shows early promise for supporting, and enhancing teachers’ learning. Research suggests that by providing regular, impartial, real-time feedback these tools can have small effects on targeted teacher practices—and do so in a cost-effective manner compared to current practice. Our Foundation is supporting several grantees using AI to improve instructional coaching and professional learning for teachers:

• FIND-PD, based out of the Stanford Center on Early Childhood, which uses AI to analyze videos to pinpoint effective practices and then delivers coaching based on those practices to upskill and empower early childhood education providers to increase responsive interactions and build child skills;
• Leading Educators, which partners with school systems on professional learning and instructional strategy, is piloting an initiative to identify scalable AI-enabled instructional practices;
• Saga Education, an evidence-based, personalized tutoring model that uses AI to identify elements of effective tutoring and deliver coaching based on these best practices;
• TalkingPoints, a messaging platform that helps schools and teachers communicate with families and utilizes AI to provide real-time mentorship that supports teachers in crafting more positive and effective communications;
• Teaching Lab, which provides curriculum-based professional learning, is curating AI tools to support teachers in implementing instructional materials with fidelity along with adapting content to individual student’s needs; and
• Teaching Matters, which provides curriculum-based professional learning to schools and is partnering with TeachFX to understand how genAI-enabled coaching can strengthen coaching sessions and ultimately teacher practices.

2. GenAI holds promise for personalizing learning through assessment

We see reasons for optimism about genAI’s potential to accelerate new and more valuable forms of student assessment, something that has long been desired in the sector. While some states have begun using genAI tools to automate grading, we’re even more excited by the potential of AI to advance assessment of individual students’ cognitive and noncognitive skills and competencies and deliver actionable, real-time insights to educators.

We’re supporting this work with leading assessment organizations as well as with curriculum providers who are increasingly integrating assessment into their products:

• ANet, which focuses on elevating student learning by enhancing assessments, curriculum, and data usage, and is launching AI-driven assessment prototypes that use predictive analytics to give educators deeper insights with less testing time;
• ASSISTments, a curriculum-connected math practice and assessment platform that uses genAI to make feedback and support more personalized and consistent for students;
• CommonLit, a comprehensive literacy program incorporating genAI and machine learning to better align assessment, professional learning, and instructional materials;
• Magpie Literacy, which offers research-based digital literacy tools for kindergarten through eighth grade and is developing AI-based differentiation that responds to student misconceptions and linguistic differences; and
• Quill.org, which provides free AI-powered literacy activities that build reading comprehension, writing, and language skills for elementary, middle, and high school students.

3. There is a need for evidence and new models to evaluate genAI’s impact

Given the infancy of this field, there’s been little research published to date on the efficacy of genAI tools in education. In a recent article, my colleague Meghan McCormick summarized the current state of evidence, pointing to emerging research on the positive impacts of AI-enabled teacher coaching products while highlighting the need for deeper understanding of genAI’s impact on students’ academic and social-emotional outcomes. Other recent studies have since raised questions about the short-term versus long-term impacts and tradeoffs genAI may present to student learning.

Guided by our belief in scaling evidence-based solutions, our Foundation has sought opportunities to build evidence on the efficacy of genAI tools. In July, we announced a $900,000 grant to the Research Partnership for Professional Learning to manage three studies focused on identifying the features and traits of genAI tools that show potential to improve teacher quality and student learning—and to disseminate those results to the education community at large.

That said, we increasingly suspect that traditional forms of research and evidence-building may prove ill-suited to study AI models and interventions that are continuously evolving. We anticipate increased importance of cross-sector benchmarks to evaluate genAI models and tools, as well as rapid-cycle research designed to increase the efficacy of product design and implementation.

4. Funders can play an important role in helping nonprofits harness genAI

As evident in the latest Stanford HAI/Project Evident poll, over half of nonprofits are beginning to integrate AI into programmatic work, while over three-quarters are deploying it for internal use cases (e.g., finance, human resources, technology, communications). This pattern mirrors what we’ve experienced with our own grantees through 1:1 conversations and a convening we hosted at ASU-GSV earlier this year to better understand AI opportunities and needs.

Based on grantee input and emerging industry trends, we are targeting three priority areas in which to offer support to our nonprofit partners:

• Leveraging AI for data analysis and internal operations;
• Developing AI policies, principles, and frameworks; and
• Creating space for shared learning about AI usage, challenges, and impact.

We will continue to expand capacity-building efforts to support the evolving genAI needs of our grantees through numerous formats including expert-led workshops, peer learning opportunities, and direct technical assistance.

5. AI in education is maturing, but the field is still fragmented

Several new initiatives and resources launched this past year indicate that genAI’s role in the education sector is growing, with increased attention on the need for policy and decision-making guidance at the district and state level. Nearly half of states have now issued guidance to support how schools use genAI. In parallel, new guidance offered by the Department of Education and the increasing momentum of the EDSAFE AI Alliance demonstrate a greater focus on bringing standards and aligned practices to this emergent field. Meanwhile, we’ve observed the launch of new AI newsletters and movements aimed at building knowledge and community.

Despite these efforts, the field remains highly fragmented with low levels of philanthropic funding. The lack of shared language, understanding, and principles limits growth and collaboration. Recognizing this challenge, we have invested in efforts to promote increased connectivity and idea-sharing across the sector:

• Center on Reinventing Public Education, in April 2024, gathered over 60 state and federal policymakers, edtech innovators, school system leaders, and advocates for the Think Forward: Learning with AI forum, which culminated in the report Wicked Opportunities: Leveraging AI to Transform Education; and
• Edtech Insiders, which is developing a new resource to track genAI-in-Ed solutions, building upon its AI Tools in Education Database and the Map of Generative AI for Education produced by Laurence Holt.

What’s Next

Taking stock of our genAI investments to date and genAI’s potential future impact on education raised numerous strategic questions for how our Foundation supports this work moving ahead. These include:

• What are the most pressing needs, opportunities, and risks we are uniquely positioned to address?
• How can we advance new and adaptive approaches to evidence-building?
• What insights and support can we deliver to grantee partners to equip them to capably and responsibly leverage genAI to scale cost-effectively?

In the months ahead we seek to develop a stronger perspective on these questions to guide our efforts in 2025. We remain committed to intentionally, collaboratively, and openly sharing what we’re learning. To stay up-to-date on the latest news from Overdeck Family Foundation, subscribe to our newsletter.

Thank you to Anthony Kuzma for your contributions to this blog post.

Header image courtesy of Quill.org