Skip to main content
As AI sweeps the education industry, teachers in one mid-sized district in California are navigating the challenges and opportunities that accompany rapid technological change. Without a district AI policy in place, educators are working through this change on their own, relying on their judgment and experience to guide classroom use. However, the policy vacuum has led to inconsistencies in how AI is applied across the district. Some teachers, already coping with immense workload fatigue and a plethora of existing EdTech tools, are reluctant to use AI. Others have responded by independently developing their own classroom rules, allowing students to engage with AI in a structured, human-centric learning environment.
AI initially arrived in the district of 20,000 students through a cautious, top-down procurement process shaped by state-level guidance. The California Department of Education encourages educators to approach AI through a lens of equity, academic integrity, and data privacy by maintaining an awareness of potential inaccuracies, biases, privacy violations, and unintended social impacts.1 The state further stresses the importance of implementing AI upon a foundation of human connection, with educator and student voices at the center. These guidelines are part of a broader national pattern where half of all states have issued AI guidance focused on preserving the human element in learning.2
At the local level, district leader Brian Donnelly describes how rollouts typically unfold in practice: “What we’ve seen in the past is that they’ll usually provide the training on the curriculum first, and then they’ll introduce how AI can help enhance what they’re doing.” He explains that this slow and steady approach increases uptake: “The key thing is rolling out the pieces one at a time and not bombarding the teachers with everything, but just focusing on core [curriculum]… then the next [step is] training.” Typical in larger districts, this kind of rollout follows a centralized procurement process in which AI adoption is a district-level decision rather than one made by individual schools or teachers, subject to top-down vetting and budget scrutiny.
In alignment with state policy, the district prioritizes academic performance, teacher usability, and data privacy when adopting new AI tools, only considering them if they meet those requirements. As Donnelly explains, “[it’s] first student achievement, and then the people who are working with this product… and then finally the data piece—if we feel the data is in a safe place, then most likely we’ll move in that direction.” This strategy creates an environment in which AI is introduced deliberately, with a focus on how students and educators will actually use it to support learning, rather than the novelty of the technology itself. As enrollment-driven public funding to the district shrinks, efficiency-promising AI tools are becoming more attractive, even as those same financial constraints threaten sustained spending on them.3
As districts like Donnelly’s work to translate broad state recommendations into day-to-day classroom decisions, they are navigating the gap between state-level guidance, district priorities, and the reality of how teachers ultimately engage with AI in practice.
Teacher-Led AI Use Varies Across Classrooms
When teachers step in to define their own relationship with AI, its applications look different from classroom to classroom. Across the district, teachers who feel empowered by the technology are eager to experiment. Donnelly describes one supplemental tool that makes life easier for both students and teachers: “Students can upload their essays, and then the AI will help score it [and] provide feedback, and teachers can do the same thing.” Dual-use tools such as these expedite student access to feedback while easing the grading burden on teachers so that they have more bandwidth for hands-on teaching. That said, teachers have to be careful about using AI as a stand-in for student interactions. In a recent poll from The Center for Democracy & Technology, 50% of students said that using AI in class makes them feel less connected to their teachers.4 Also, while AI may be able to deliver feedback faster than a busy teacher, it risks being low-quality or incorrect, diminishing its potential as an educational tool. This is one area where district oversight is crucial to preserving the human connections that promote engaged learning.
For Michael Sutton, another teacher in the district, certain AI tools provide targeted data to inform classroom instruction. For example, his students have been benefiting from learning accelerators like Microsoft Reading Progress, which tracks their reading fluency in real time. “[AI] is really awesome,” he says, “because [it] allows them to record visually and audibly… and it’ll tell you words that are omitted, inserted, words-per-minute, accuracy… a whole gamut of data… The AI stores it for you, and it’s there available for the teacher.” These tools are providing highly detailed insights into student progress that teachers simply didn’t have access to before. While these tools can help teachers better tailor learning to their students, it’s worth noting that they can also add to teacher workload in ways that may not be addressed in district-level training or support, potentially shaping uneven instruction across classrooms.
Beyond data-driven instruction, teachers are also discovering ways to incorporate AI into creative student-centered classroom activities. Sutton describes how he used AI to upgrade his biology lessons, allowing his students to script and record unique rap battles about cell organelles using generative tools: “Now with AI, my students were creating their versions of the cell and also creating new songs… they were taking ownership of it.” To ensure students were actually learning from the project, Sutton built an independent assessment into the lesson to verify understanding, asking his students, “Do you know what this organelle does… can you define it, and can you tell me its function?” Sutton’s approach illustrates how educators can pair students’ creative AI use with direct verification of understanding to prevent overreliance on generative tools.
Outside of instruction, teachers like Mercer have created AI assistants to help with various administrative jobs. For example, Google Gemini’s popular “Gems” feature allows users to build custom AI experts with specific instructions to help with repeatable tasks, such as lesson planning or paper grading. “With Google Gemini, I can really customize it how I want,” she says. “I can have a Gem for my administrative work… I could have my Gem as a teacher collaborator, or I can make that certain Gem for data pulling.” Whether it’s a specialized grading platform or a fine-tuned AI assistant, these tools allow teachers to maintain control over their workflows while freeing up time for high-level instruction.
Of course, not every teacher is eager to embrace AI in the same way. For some, the absence of guardrails has done more to deepen psychological discomfort than invite practical experimentation. Veteran teachers, in particular, worry that adopting AI means relinquishing some measure of control over their teaching.
Teachers Are Drawing A Line Around Human Value
Among teachers who are hesitant about AI, many fear that handing over work to generative tools means surrendering their professional expertise, a notion that threatens their sense of autonomy and professional identity. “I think the challenge comes from the mindset… [that] now you’re giving up that control or you’re giving up your expertise in a particular area,” Donnelly says. That loss can feel especially threatening for teachers who have spent decades refining their workflows. Mercer relates to teachers who feel this way: “You often hear the jokes about how teachers and people who work in schools value control. If you’ve been teaching for twenty-some years… it’s hard to give that up and trust machine software to do that for you.”
For district leaders, addressing the psychological barriers to adoption means showing teachers how AI can give them more control over their instructional capacity as they grow accustomed to working with supplemental tools. Donnelly likens AI to the emergence of automated driving: “If you sit behind the automated vehicle, it drives for you. We’ve been driving for thirty-something years, and there’s a comfort of steering the wheel… but when somebody else does it, it’s uncomfortable at first, and you have to get used to it.” Yet, even in self-driving vehicles, human judgment and oversight [are] essential, just as teachers still need to maintain control in the classroom. For teachers who are willing to experiment with AI, that initial sense of discomfort eventually gives way to collaboration, where teachers extract real value from automated processes while preserving autonomy over their work.
Naturally, that adjustment has its limits. As Mercer explains, “children are still developing, and the way they develop hasn’t changed much, but what they’re exposed to has.” Teachers still need to exercise judgment regarding how and when to use AI with younger students. Educators in the district are keen to establish strict human boundaries, walling off early childhood education—notably Transitional Kindergarten and Kindergarten—from AI to ensure children continue to receive the human interaction that is crucial during these crucial developmental years. While Sutton feels that AI has been an excellent addition to his classroom, he stresses that the technology should be limited to the older grades. “I feel these youngsters need all the interaction and development that they can get human-wise,” he says. “Don’t give them a laptop to read out loud or an AI to read them a book. You teacher, you TA [Teaching Assistant], you read [to] them. They need that.”
This emphasis on human interaction also applies to Social-Emotional Learning (SEL), which relies heavily on interpersonal connections and relationship-building: areas where educators say AI involvement should stay minimal or absent. Teacher Maria Reyes reiterates that AI is not an adequate stand-in for “the human connection… the teacher getting to know her students.” She adds that “the T-SEL [Transformative Social and Emotional Learning] component is definitely [where] the teacher cannot be replaced.” Regardless of age, students need human connection throughout their schooling. These perspectives echo that of the California Department of Education, which emphasizes that technology should be used to enhance rather than replace the human touch.1 The state cautions educational leaders against overreliance on technology that dilutes essential human elements like teacher interaction and mental health supports.
A key challenge for the district is supporting teachers as they balance rapid innovation with this irreplaceable human element. Mercer describes how most teachers see it: “They realize, yes, [AI] is here. It’s a great collaborator. However, there’s nothing that’s going to replace the human.” Some AI tools already exist to facilitate this human-centered approach. Microsoft Reflect, for example, is a well-being application that uses emotional temperature checks to help students build emotional self-awareness. At the same time, the technology surfaces crucial insight into student emotional needs, empowering teachers to step in for the follow-through and ensure support remains entirely human.
That same people-first approach also applies to teachers, many of whom are already feeling spread thin by workload fatigue and new technology.5 Reyes shares the words of one of her colleagues: “I don’t want more work, I don’t want to do anything new. Just show me how I can be efficient and make things that are interesting and relevant to my kids. Make my job easier, because I have enough on my plate.” As outlined in state AI guidance, AI should reduce administrative burden so that teachers are free to focus on the deeply human aspects of teaching.1 For teachers who are struggling under heavy workloads, that promise requires clear district policy: guidance that shows them how AI can streamline their work and allow more time for connecting with students.
Future-Ready AI Ambitions Outpace District Quality Control
The California district is balancing two key priorities: protecting childhood development by preserving foundational learning in early education and supporting older students as they engage with the generative tools they’ll be expected to use in postsecondary education and future careers. The focus, Sutton says, is on teaching students the skills they will need to remain competitive when they enter the workforce: “It goes back to training, knowing why you’re using this tool…What do you want to do with this tool? What’s your plan? What’s your outcome? What standards are you trying to hit?” Enthusiasm toward AI is growing as more teachers acknowledge the necessity of AI fluency after graduation, especially when tools demonstrate improvements in student learning while still preserving teacher input and human connection.
Many educators view the rapid pace of technology as an urgent mandate to get students up to speed. Donnelly explains that much of this forward momentum is being championed by teachers who see the value of AI, actively attending conferences and laying groundwork in the classroom to prepare students for the future. Sutton is one such educator. “Some students next year, they’re going to be out there competing in this world,” he says. “Our job is to prepare our students for the real world that’s coming up.” For Sutton, that means if a tool exists that can help students build future-ready skills for college and the workforce, they must engage with it.
By contrast, tools that aren’t meeting this standard need to be filtered out. Mercer puts it simply: “If these AI tools are not going to be teaching our students sustainable future-ready skills for them to be successful… no matter if they go to a trade school or a community college or a four-year or directly into the workforce, then I would say it’s not very valuable.” Yet, districts lack access to evaluation tools to translate these priorities into actionable criteria, leaving much of the evaluation work up to individual judgment.
While the district approaches the adoption of new AI tools with strict attention to student performance and future preparedness, exactly how teachers use AI for their own purposes remains largely unchecked. A survey from the EdSignals Studio found that 66% of teachers reported modifying or adapting existing materials using AI, and this was the single most common use of AI across core curriculum, EdTech, and professional learning materials.6 Yet, little is known about how teachers are modifying materials and to what extent it’s affecting the quality of instructional content.
Without comprehensive guidance from the district, individual teachers must determine how AI-modified materials meet instructional expectations. This can result in uneven practices across classrooms and uneven learning outcomes for students. More oversight is needed to ensure modified materials meet the same quality standards as the original materials. Closing that gap also requires giving teachers a clear way to check their work, such as a rubric that lets them measure their AI-modified or generated materials against the same metrics used to evaluate other components of the curriculum. With district funding at risk, modified materials may become more commonplace as teachers are pushed to do more with less. While funding pressures might limit the amount of oversight the district can provide, a human-centric approach—one that prioritizes teacher judgment while providing clear guidance for the responsible use of AI—can help ensure these tools consistently support student success and equitable access to future-ready skills.
Closing Gaps in Human-Centered AI
At every grade level, the district’s AI strategy is focused on using AI to support learning and future readiness while maintaining the crucial human connection that fosters interpersonal skills and emotional intelligence. However, the absence of clear direction has left educators in uncharted territory, forcing individual teachers to decide how AI should be used in their classrooms. Many teachers in the district are modeling what balanced AI use looks like in practice, letting students engage creatively while ensuring that real learning is taking place. Others are reluctant to take on new tools in the face of existing workload pressures, with some choosing to avoid AI altogether.
The opportunity lies in using AI to make teachers’ lives easier, preserve their individual sense of control, and build confidence that the materials they’re implementing will meet educational standards and strengthen their connection with students. As teachers across the district experiment with AI, human-centric guidance will help preserve teacher autonomy and support student development, ensuring classroom practice aligns with the district’s future-looking vision.
Names in this article have been changed to preserve anonymity.
Sources
- California Department of Education. (2025). Guidance for the Safe and Effective Use of Artificial Intelligence in California Public Schools. https://www.cde.ca.gov/ci/pl/aiincalifornia.asp
- AI for Education. (2025, October 28). State AI guidance for Education. AI for Education. https://www.aiforeducation.io/ai-resources/state-ai-guidance
- Hudson, E. (2026, April 16). Schools Bet on AI Tools Like CoPilot, Gemini to Bolster Finances. Bloomberg. https://www.bloomberg.com/news/articles/2026-04-16/us-public-schools-wager-on-ai-to-tackle-financial-woes
- Laird, E., Dwyer, M., & Quay-de la Vallee, H. (2025, October). Hand in Hand: Schools’ Embrace of AI Connected to Increased Risks to Students. Center for Democracy & Technology. https://cdt.org/wp-content/uploads/2025/10/FINAL-CDT-2025-Hand-in-Hand-Polling-100225-accessible.pdf
- Klein, A. (2022, March 8). Tech Fatigue Is Real for Teachers and Students. Here’s How to Ease the Burden. Education Week. https://www.edweek.org/technology/tech-fatigue-is-real-for-teachers-and-students-heres-how-to-ease-the-burden/2022/03
- EdSignals Studio, Ecosystem Study #2, 2026.
