AI to retain teachers: when the algorithm acts as a critical mirror for school principals

By Zendoric · June 30, 2026.

Orlando hosted ISTELive26 this week, the major U.S. education-technology fair, and among the usual demos of chatbots for students a less obvious angle emerged: using AI as a strategic-thinking tool for school administrators themselves, not for their classrooms. Nick Schiner, director of peer learning at Digital Promise —an organization that works with Google Education and district leaders—, presented a practical framework for principals and superintendents to leverage language models in tackling teacher retention, the most pressing human-resources problem in the U.S. K-12 system.

Schiner's method has a logic closer to consulting than to pedagogy. His first principle: build a rich 'context block' before asking any question. Job title, geographic location, number of students, public academic performance, ongoing initiatives. That last point produced one of the most revealing findings he recounted: when principals dumped all their initiatives into the chatbot, the AI handed back something no subordinate had dared tell them —that the accumulation of simultaneous projects could itself be a cause of teacher burnout.

The second principle is counterintuitive: explicitly instruct the model *not to offer solutions*. Instead, ask for research synthesis, identification of gaps in the argument, counterarguments. Schiner frames it as protection against the premature delegation of judgment: the administrator must reach their own conclusions; the model should accompany that process. The third move —asking the chatbot to play 'devil's advocate'— fills a real void in hierarchical organizations: leadership teams rarely receive frank criticism from their subordinates. A model with no employment relationship and no fear of being fired can give the uncomfortable feedback no one in the room wants to articulate.

Perhaps the most concrete and valuable example Schiner offered was the fourth principle: asking the model "what or who am I missing?". Several principals he worked with discovered this way that their mentoring programs were designed for new or struggling teachers, but systematically ignored the strongest teachers —the very ones later asked to serve as mentors. The question that emerged was not new, but no one had framed it with such clarity: are we burning out our best teachers precisely because we consider them indispensable?

**Our reading: the classroom is not the only front**

The most valuable thing about this proposal is not the five tips themselves —they are, in essence, good prompting practices— but what they signal about how AI use is maturing in the education sector. For years, the debate focused almost exclusively on the impact on student learning: will students cheat? will they learn more? less? That is a legitimate discussion, but a partial one. Educational organizations have a second tier of problems —talent management, strategic planning, resource allocation— that rarely receives technological attention, partly because school principals usually have no access to outside consultants or analytics departments.

AI as a 'thinking partner' for managers without a team fills that gap cheaply and at scale. The principal of a rural school in a state with chronic teacher shortages cannot hire McKinsey; they can open a chatbot with the right context and subject their hypotheses to systematic scrutiny. This is not trivial. It is a democratization of strategic analysis that until now was reserved for large urban districts with resources.

But it is worth keeping in mind the warning implicit in Schiner's fifth principle —practicing the same AI literacy demanded of students— because the opposite risk is real. The comfort of receiving a structured analysis can lead principals untrained in critical prompting to adopt the model's output without questioning it. And here the irony is considerable: the Education Week article itself is flanked by news about the resignation of LAUSD superintendent Alberto Carvalho, under the shadow of a failed student chatbot initiative and an FBI investigation. The lesson is not that AI doesn't work in education; it is that governance matters as much as the technology.

In Zendoric's analysis of employment in education we have argued that the teacher who wins in the AI era is the one who orchestrates the tool, not the one who fears it nor the one who blindly delegates to it. The same applies to the school administrator: AI is not going to solve the teacher-retention crisis on its own, nor will it write a staffing strategy that works without someone who knows the real dynamics of each school. What it can do is act as an analytical colleague available at three in the morning, with no political agenda and no fear of pointing out what the leadership team would rather not hear. That, used well, is a genuine asset.

In the short term, the temptation to replace human judgment with the comfort of automated output is a concrete risk, especially in institutions with little culture of critical thinking or under pressure for immediate results. In the long term, however, the possibility that every school principal could have access to quality strategic analysis —regardless of the size of their district or their budget— points to exactly the kind of reduction in structural inequality that AI has the potential to produce when applied rigorously.