Generative AI use in educational settings is accelerating at a rapid pace, with students and teachers across the country utilizing and experimenting with educational technology embedded with AI capabilities or with Large Language Models (LLMs) themselves. 

For most school districts, navigating this nascent space is an enormous challenge for educators, students, communities, and policymakers. Some of these complexities are operational: What are the goals and ideal outcomes framing AI procurement and implementation? How does a district decide what AI tools to procure? What data security and privacy risks might AI integration introduce? What role do, and should, parents, students, and community members play in K-12 AI procurement and implementation strategies? Are there specific tools or AI use-cases that have shown promise, and/or what should be avoided?

Upskilling of staff, students, and leaders is another challenge of AI implementation with considerations including: How do we train educators, school leaders, and district leaders about responsible and research-backed strategies to safely integrate AI into day-to-day administrative practices or, more importantly, into classroom instruction that directly impacts students? Who designs, and how do they design, AI literacy trainings for our education stakeholders? How do we ensure AI literacy is accessible for districts and schools? And, some questions are more philosophical about the role of AI in education broadly: Is AI “good” for education and who decides what good AI use looks like? 

A recent survey from the Center for Democracy and Technology found that 86% of students and 85% of educators have experimented with AI technologies during the 2024-25 school year — simply put, the cat is out of the bag. If students are using AI, teachers are using AI, and system leaders are using AI, then AI is already a part of our education system whether we want it to be or not. The equity impacts could be immense, positive and negative, necessitating a real need for the policy and research ecosystem to invest time and energy into this space. 

Advancement of research and policy needs to be the number one priority to ensure responsible and equitable integration of AI into our nation’s education systems —  it is critical that this work does not happen in a vacuum. To accomplish this aim, the field first needs to get smarter about how schools and districts, especially those that serve high populations of students of color and students from low-income backgrounds, are navigating the procurement, implementation, and evaluation of AI in their educational settings. 

Currently, EdTrust is leading a research project to help contribute to this need. We are conducting focus groups with teachers, school leaders, and district leaders in under-resourced communities — sitting down directly with these stakeholders to uncover: 

• Current adoption stages: Where do schools and districts actually find themselves in the process of AI implementation and skill-building? 
• Resource needs: What specific questions do stakeholders have, and what supports are they missing? 
• Hopes and worries: How are folks navigating the tension between the promises of AI integration (saved time, potential student outcome gains, etc.) and the perils (data privacy, systemic bias, misuse and over-reliance, offsourcing critical thinking, exacerbation of existing disparity gaps, etc.)? 
• Navigating biases: How are leaders and teachers thinking about the inherent biases of LLMs, and how and if they are addressing this issue with students and school staff, if possible? 

At the end of the day, equitable implementation of technology is a purposeful choice. Powerful technologies do not inherently lead to more equitable conditions for marginalized student populations; our decisions and actions as policy leaders, superintendents, principals, teachers, parents, students, and civil rights organizations must be intentionally designed to foster equity and fairness. 

Earlier I asked: Who decides what good AI use looks like? The answer is us. But successfully answering this question, ensuring that AI integration into educational systems lifts all students rather than widening existing disparity gaps, requires us all to anchor our policies and practices in the needs, hopes, and concerns of the educators and leaders serving America’s most marginalized students. To do otherwise, to activate autopilot and hope for the best, will inevitably lead to AI becoming the next digital divide that vulnerable student populations, and the policy ecosystem that serves these students, will have to fight to overcome.