The generative AI revolution promised to lighten the load for overworked educators, but the reality has been far more complex. For teachers already stretched thin by lesson planning, grading, and administrative duties, the sudden emergence of powerful chatbots like ChatGPT has introduced a new layer of pressure: the demand to master a technology that seems to evolve weekly. Many feel stranded without a clear roadmap, caught between institutional encouragement to adopt AI and a lack of concrete, practical guidance. This tension is playing out in schools across the country, but a case study from Connecticut offers a nuanced, pragmatic alternative to the chaos.
The AI Burden: From Time-Saving Promise to New Pressure on Teachers
The narrative around generative AI in education has been dominated by its potential to save time. Automate grading. Generate lesson plans. Personalize feedback. For teachers who routinely work nights and weekends, these promises are seductive. Yet the implementation has been anything but seamless. The boom in generative AI tools has paradoxically increased the burden on educators, who must now navigate a confusing landscape of platforms, policies, and ethical considerations on top of their existing responsibilities.
Organizations as influential as OpenAI and UNESCO have publicly encouraged AI integration in classrooms, framing it as inevitable and beneficial. But on the ground, teachers report feeling confused about how to proceed. An EdWeek report from January 2025 captured this sentiment, noting that school policies remain opaque and inconsistent. The gap between high-level endorsement and classroom-level execution is vast, and it is teachers who are left to bridge it.
Why Teachers Are Struggling to Find a Clear Path Forward
The confusion is not simply about which tool to use. It is about purpose, ethics, and pedagogical fit. Teachers are expected to evaluate AI tools for accuracy, bias, privacy compliance, and educational value — all while maintaining their regular workload. The technology itself is imperfect: it can hallucinate facts, reinforce stereotypes, and generate inappropriate content. Without clear institutional guidelines, many educators either avoid AI entirely or adopt it haphazardly, creating inconsistency across classrooms and departments.
The question that follows naturally is: how can a school approach AI training in a way that is both practical and principled? One answer comes from a small private school in Connecticut that has been quietly developing a model that prioritizes flexibility over mandates.
Inside Cheshire Academy: A School That Chose Training Over Mandates
Cheshire Academy, a private boarding and day school in Connecticut serving approximately 400 students in grades 9 through 12, has taken a notably different approach. Administrators there do not force instructors to use AI. Instead, they have invested in training that empowers teachers to make informed choices. George Aiello, the school’s librarian and technology coordinator, reports that the “vast majority” of instructors now use AI in some capacity — a testament to the effectiveness of a voluntary, education-first strategy.
What makes Cheshire Academy’s approach instructive is not the specific tools they use, but the philosophy behind their training. The school did not prescribe a single platform or mandate usage quotas. Instead, on the advice of external consultants, they focused on general techniques for working with AI. This included how to craft effective prompts, how to evaluate outputs critically, and — crucially — how to recognize the technology’s limitations, particularly its potential for generating incorrect or biased responses.
The Patchwork of Tools at Cheshire Academy
The educators at Cheshire Academy use a variety of tools, reflecting the school’s laissez-faire adoption model. General-purpose chatbots like ChatGPT and Perplexity are popular, as is MagicSchool, an AI-powered platform designed specifically for educators. This patchwork approach is not a sign of disorganization; it is a deliberate strategy. By allowing teachers to choose tools that fit their subject area, teaching style, and comfort level, the school fosters organic adoption rather than top-down compliance.
MagicSchool, for instance, offers features tailored to common teaching tasks: generating quiz questions, drafting email communications to parents, and creating differentiated reading passages. ChatGPT and Perplexity, by contrast, are more open-ended, suitable for brainstorming lesson ideas or summarizing complex topics. The diversity of tools means that teachers can experiment and find what works for them, rather than being forced into a one-size-fits-all solution.
How Teachers Are Actually Using AI: Lesson Plans, Rubrics, and the Feedback Frontier
At Cheshire Academy, the most common use of generative AI is in the preparation phase of teaching. Teachers regularly ask AI tools for help in planning lessons and creating grading rubrics. This is a low-stakes, high-reward application: AI can generate multiple variations of a lesson plan on a given topic, suggest alternative activities, or produce a rubric with clear criteria for assessment. Teachers can then adapt and refine the output, saving hours of upfront work.
This use case aligns with broader trends in education technology. Lesson planning is a time-intensive task that benefits from the kind of rapid prototyping AI enables. A teacher covering the French Revolution, for example, can ask ChatGPT to generate a timeline of key events, a set of discussion questions, and a list of primary sources — all in minutes. The teacher then curates and customizes, maintaining control over quality and relevance.
How does a teacher use AI to create a grading rubric? A teacher can prompt a tool like MagicSchool or ChatGPT with the assignment’s learning objectives, grading criteria, and desired point distribution. The AI produces a structured rubric with categories such as “Analysis,” “Evidence,” and “Clarity,” each with descriptors for different performance levels. The teacher reviews the rubric, adjusts language, and aligns it with classroom standards before use. This process reduces drafting time from an hour to under ten minutes while maintaining pedagogical rigor.
The Feedback Gap: Why Quality, Personalization, and Privacy Are Holding Teachers Back
One area where Cheshire Academy teachers have not yet fully deployed AI is in giving feedback to students. Some educators are interested in using AI to provide initial comments on student work, but concerns over quality, personalization, and privacy have prevented implementation. This caution is well-founded. Automated feedback can miss the nuance of a student’s argument, fail to account for individual learning trajectories, or inadvertently expose sensitive student data to third-party servers.
The privacy concern is particularly acute for schools. Student work may contain personal reflections, identifiable information, or proprietary assessment data. Uploading this material to a cloud-based AI tool raises questions about data storage, retention, and compliance with laws like FERPA in the United States. Until these issues are resolved at a policy level, many schools will rightly hesitate to use AI for direct student-facing feedback.
The Teacher Who Doesn’t Use AI: Miriam Przybyla-Baum and the Value of Experience
Not every teacher at Cheshire Academy has embraced AI for their own workflow. Miriam Przybyla-Baum, who teaches French, is a notable example. With nearly 30 years of teaching experience, she has accumulated a rich repository of classroom materials. She simply does not need AI to help her plan lessons or create assignments. Her expertise allows her to draw on proven activities, readings, and assessments without the aid of a chatbot.
Yet even Przybyla-Baum engages with AI in her teaching — because her students are already using it. She first noticed students taking shortcuts with AI-powered tools like Google Translate years before ChatGPT was launched. The technology was already reshaping student behavior, and she had to adapt her teaching to address it. For language teachers, the challenge is acute: if a student can instantly translate a paragraph or generate a grammatically perfect sentence, how do you assess genuine language acquisition?
Przybyla-Baum’s approach is to address AI directly in her classroom. She talks with students about when translation tools are appropriate and when they undermine learning. She designs assessments that require live demonstration of skills — oral exams, in-class writing, collaborative tasks — that are harder to automate. Her experience highlights an important truth: even teachers who do not use AI professionally must still contend with its presence in their students’ lives.
What Cheshire Academy’s Model Teaches Us About AI in Schools
The Cheshire Academy case offers several lessons for schools grappling with AI integration. First, voluntary adoption combined with high-quality training is more effective than mandates. Teachers who feel pressured to use AI often resist or use it superficially. Those who are educated about its capabilities and limitations — and then given the freedom to choose — are more likely to adopt it thoughtfully and effectively.
Second, training should focus on principles, not products. By teaching general techniques like prompt engineering and critical evaluation of outputs, Cheshire Academy prepared teachers to use any AI tool, not just a specific one. This is crucial in a landscape where tools change rapidly. A teacher who understands how to craft a good prompt can apply that skill across ChatGPT, Claude, Gemini, or a specialized education platform.
Third, schools must respect the expertise of experienced teachers. Miriam Przybyla-Baum’s decision not to use AI for lesson planning is not a failure of adoption; it is a rational choice based on decades of accumulated resources. AI is a tool, not a replacement for professional judgment. Schools should celebrate teachers who use AI effectively, but they should also respect those who choose not to — as long as they remain aware of how their students are using it.
What Are the Risks of Using AI in Education?
The primary risks are threefold: accuracy, bias, and privacy. AI models can generate incorrect information with confidence, which is particularly dangerous in educational contexts where students are building foundational knowledge. Bias in training data can lead to outputs that reinforce stereotypes or exclude marginalized perspectives. And privacy concerns around student data remain unresolved, especially when using free, consumer-grade chatbots that may not comply with educational data protection laws. Schools must address all three risks before scaling AI use.
Strategic Implications for Schools and Districts
The Cheshire Academy model is not a one-size-fits-all solution, but it points toward a strategic approach that other schools can adapt. The key elements are: invest in teacher training before deploying tools, emphasize critical thinking about AI outputs, allow choice in tool selection, and establish clear ethical boundaries — especially around student data and feedback.
For districts with limited budgets, this approach is actually more cost-effective than buying expensive, enterprise-grade AI platforms. Training can be delivered in-house or through low-cost external consultants. Free or low-cost tools like ChatGPT and MagicSchool can serve most needs. The real investment is in time: giving teachers the space to learn, experiment, and reflect on their practice.
Looking ahead, the schools that will thrive are those that treat AI not as a shortcut but as a collaborator. The goal is not to automate teaching but to augment it — freeing teachers to focus on the human elements of education: mentoring, discussion, and personalized support. Cheshire Academy has taken a step in that direction, and its experience offers a grounded, realistic blueprint for others to follow.
