Developing an R&D Strategy for AI Integration and School Transformation
Plus my NAEP take, new tech developments, and new research
I’m sure many of you are diving into yesterday’s NAEP score release. You can peruse my thoughts on the scores in The 74 this morning (including on some figures that deserve more attention), but I think many of us are in agreement about the key takeaway: American students are receiving neither the educational opportunities they deserve nor those that will enable them to thrive in an increasingly complex society and changing economy.
These results tell me that the need for dramatic reform has never been more urgent—and this reform must include investment in emerging technologies. AI-driven tools to support personalized learning, tutoring, curriculum, and assessment can help ensure all students get the support they need while empowering educators to be more effective.
Tech Developments (Lots of Them!)
The best explainers I’ve seen on the DeepSeek Freak were from the Hard Fork podcast and Claire Zao’s newsletter. It’s worth taking some time to learn the nuance and basic terminology of these “foundation models,” which train on massive datasets so they can have multiple use cases. LLMs, for example, are a type of foundation model. The AI market is shifting so fast and in such profound ways, affecting global markets and potentially global security.
With all the attention on DeepSeek, you may have missed the debut of Operator, an AI “web agent” that you can get by paying for OpenAI Pro. Operator connects to controls in your computer to search the web and do innocuous and potentially helpful things like book tickets to a concert or make restaurant reservations. But obviously there is also potential for web agents to be used for nefarious purposes, even with safeguards in place. If you read my review of Nexus you may recall the OpenAI experiment to see if AI could get through CAPTCHA tests (spoiler alert: it did).
Speaking of AI agents and big existential questions… this session at the World Economic Forum was fascinating. Highly recommend. Here’s the summary:
Artificial general intelligence could possess the versatility to reason, learn, and innovate in any task. But with rising concerns about job losses, surveillance, and deepfakes, will AGI be a force for progress or a threat to the very fabric of humanity?
Reimagining Education with Emergent Tech: What’s the Right Approach?
A recent BCG report makes the case for disruptive innovation in education using emergent technologies like AI, arguing that “tomorrow’s education must be personalized, experiential, adaptable, scalable, inclusive, and sustainable.” The report lays out a long list of ways technology could be used to transform education. The BCG team laid out a nice typology of before-, during-, and after-school tools with examples of use cases.
Although I am highly sympathetic to the above principles and I do believe that AI can help enable those things, the report’s vision for what a student’s daily “Journey of the Future” could look like seemed fairly unmoored from what we know matters to student success: strong relationships with both adults and other young people, really engaging tasks, coherent evidence-based instruction, high expectations, and high support.
Our field desperately needs an R&D program focused on how GenAI and other technologies could truly transform education in ways that are grounded in what we know kids need to succeed, what students and families say they want, and what makes schools highly effective.
At CRPE, we’ve been giving some thought to the question of a structured R&D strategy for pairing AI with coherent school transformation. We recently sketched out a set of ideas and potential investments, shared below. We’re curious to know your thoughts on this. What’s missing? What’s low-priority? Add a comment here or shoot me a note.
How to Experiment and Scale an AI-integrated School Approach: A Working Theory of Action
1. Start with “Transformation School Pilots”
Rather than creating entirely new schools, consider piloting AI-driven transformations within existing ones. This could involve:
Identifying a few willing districts or schools with strong leadership buy-in.
Funding integrated AI implementation like curriculum, staffing, and personalized learning simultaneously.
Funding networks of schools working on specific educational challenges that require integrated approaches, like career pathways or reimagining special ed.
Partnering with ed-tech providers (e.g., Khan Academy, Microsoft, Google) to create structured AI implementation frameworks.
2. Focus on Leadership Capacity-Building
If we're currently at the teacher innovation stage, school leaders must be the next focus. Equip them with tools and frameworks to lead AI-driven change, such as:
AI leadership academies for principals and superintendents to inspire and prepare promising leaders to use AI to try coherent improvement strategies
Cohort-based learning models where “early adopter” school and district leaders share best practices and challenges.
3. Integrate Policy and Reporting
AI’s potential to streamline operations (e.g., attendance tracking, intervention tracking, personalized education plans) could be a major driver for adoption at the district level. Working with policymakers to define new AI-powered data and operations systems can speed up acceptance and scale. Solving for district-level challenges like integrated data systems, more nimble procurement systems, hiring and recruitment, and privacy and bias issues can clear the path for more widespread school-level adoption as well.
4. Incubate Future-Ready AI School Designs
Collaborating with accelerators and ed-tech hubs could provide schools with agile experimentation environments without bureaucratic constraints. These spaces could serve as “sandbox environments” where educators co-develop AI strategies that are later scaled. Such experimentation hubs must be grounded in the problems schools most need to solve and should focus on coherent systems of AI-powered solutions, not just tools.
Bring together a “Tiger Team” of high-quality school and network leaders, innovative thinkers, and tech leaders to do prototype designs for AI-powered and future-ready school designs, grounded in a vision of the future that is about more joyful, equitable, rigorous, engaging, and relevant schools.
Provide support and incentivize a new crop of ed entrepreneurs (and attract back the charter management organization (CMO) leaders who left the sector) to start and scale new AI-powered school models, including microschools and other nimble delivery models. These schools could also lead the way in AI student literacy and career prep at the HS level.
5. Design for Scalability Through Networks and Case Studies
Once successful models emerge from pilot schools, create blueprints and toolkits to replicate their success in different contexts. Leveraging existing school networks (e.g., charter networks, large public districts, independent schools) can accelerate adoption.
6. Establish an AI-Education Research Consortium
Creating a collaborative ecosystem between schools, researchers, ed-tech companies, and policymakers can help systematically collect and analyze data. This consortium could:
Conduct rapid cycle evaluations to track student achievement, engagement, and socio-emotional outcomes.
Use mixed methods research to study and build knowledge on implementation challenges and solutions, policy, and politics.
Provide guidance to school and systems leaders to assess the efficacy of tools and integrated strategies.
Provide guidance to ed-tech leaders about system barriers and solutions.
Compare AI-enhanced schools with one-off tools to measure effectiveness.
Explore ethical considerations, such as AI bias and data privacy, ensuring fair implementation across diverse demographics.
Provide a centralized repository of AI education case studies to share insights across schools.
7. Build an Evidence-to-Implementation Pipeline
Once research identifies promising AI-driven school models, structured pathways for scaling should be established. This includes:
Developing implementation toolkits based on proven success stories to guide new schools.
Hosting learning communities where school and district leaders can share experiences and best practices.
Disseminate learning to government, philanthropies, policymakers, and others about evidence-based AI integration in schools.
Research News
Our friends over at Stanford’s education policy and research initiative within the Stanford Accelerator for Learning just launched a new initiative to support research and learning. SCALE will encompass existing projects like the National Student Support Accelerator, Tips by Text, and Getting Down to Facts, as well as the newly launched GenAI for Education Hub, all aimed at advancing equitable education solutions.
The GenAI Hub includes a great “repository” for research on GenAI in education. We love working with Susanna and her team and know they will produce great work in years to come.
In a new blog post for CRPE, Morgan Polikoff and his colleagues at USC explore the question: What do parents know about Generative AI in schools? The team gives us four big takeaways:
Schools and teachers aren’t communicating with parents about GenAI.
Parents mostly either don’t know whether their children are using GenAI or think their children are not using it (see graph below).
Parents have mixed and hesitant views about the potential role of GenAI in education.
There are gaps between more and less educated parents in their views on GenAI, with more educated parents generally more supportive.
Morgan’s post is the first in a series we’re running on our Lens blog highlighting expert voices on AI. Watch this space for more!
Final Words
