ADMIN SURVIVOR

The first sign that artificial intelligence had truly arrived in schools was not a revolution in learning. It was a calendar invite titled “AI Conversation — Urgent/Optional.” The invite included seven administrators, three instructional coaches, one technology director, a board member who had read half an op-ed, and a vendor whose title contained both “solutions” and “futurist.”

By 8:17 a.m., the staff had sorted itself into the traditional innovation positions: excited, suspicious, exhausted, technically curious, and “can we please finish the evaluation cycle first?” Students, meanwhile, had already become prompt engineers with the quiet confidence of raccoons operating a vending machine.

The essays changed first. Suddenly, seventh graders were deploying semicolons with the precision of Victorian barristers. Reflections contained phrases like “multifaceted pedagogical journey.” One student referred to recess as “unstructured kinesthetic peer collaboration.” He had been absent that day.

The district response was swift and technically impressive: a committee was formed. The committee produced a draft policy, revised it, tabled it, revived it, sent it to legal, and then asked an AI tool to summarize it because no human being could survive paragraph four.

The problem was not whether AI belonged in schools. The problem was that it had already enrolled, found the copier code, and joined three Google Classrooms under the name “Not A Robot.” Administrators were asked to champion innovation while also identifying cheating, protecting data, evaluating vendors, calming parents, training staff, and somehow keeping lunch from becoming a decentralized autonomous organization.

By Friday, the official guidance was clear: teachers should use professional judgment, students should use ethical judgment, parents should use patience, vendors should use fewer adjectives, and everyone should stop pretending the “AI detector” was more than a Magic 8‑Ball wearing a blazer.

At first, the lesson plan looked promising. It had a hook, a mini-lesson, guided practice, independent practice, formative assessment, vocabulary, differentiation, accommodations, enrichment, closure, SEL integration, and a reflective journaling component for the stapler.

Then the teacher noticed the standard code. It did not belong to the state, the subject, or any known democratic institution. The chatbot had aligned the lesson to “ELA.7.AI.42b: Students will analyze thematic resonance through collaborative empathy mapping and moderate jazz.”

The activity required 90 minutes, six stations, laminated cards, an anchor chart, three colors of sticky notes, and “one adult facilitator per four students.” The teacher had 38 minutes, a dry erase marker with emotional damage, and a fire drill.

Still, there was something seductive about the machine’s confidence. It never said “I don’t know.” It said, “Certainly.” It said, “Here is a comprehensive plan.” It said, “Have students create a podcast, infographic, debate, poem, and peer-reviewed article by Friday.”

In the end, the teacher did what teachers have done since chalk was considered disruptive technology: took the one usable idea, ignored the rest, rewrote the directions, and made it work with half the time and none of the promised materials.

The AI detector entered the building with a dashboard, a confidence score, and the serene moral authority of a coin toss. Teachers uploaded essays. The detector spun. The quiet student who reads college-level novels was flagged as “97% AI.” The copied paragraph from the first search result came back “Likely Human.”

Administrators were told to avoid relying on detection tools as sole evidence, which is policy language for “please do not hand a lawsuit a microphone.” The tool was useful, everyone agreed, as one data point among many, provided the data point wore a tiny clown hat.

Appeals developed their own ritual. A student explained their writing process. A parent forwarded four articles. A teacher produced drafts. Someone asked whether Grammarly counted. Someone else asked whether spellcheck counted. The assistant principal stared into the middle distance and briefly envied cafeteria trays.

By the end of the quarter, the school adopted a more reliable method: know your students, require drafts, use oral defense when appropriate, design better assignments, and treat detector scores like weather forecasts from a haunted barometer.

Prompt engineering, according to the training, is the skill of asking a machine for exactly what you want. School leadership, according to reality, is the skill of asking thirteen people for exactly what you want and receiving a spreadsheet, a concern, and a laminated acronym.

Useful prompt: “Write an email that says no without using the word no, references student safety, thanks the sender for their partnership, and does not create a precedent that will ruin April.”

Advanced prompt: “Summarize this 83-page mandate into five bullets, three risks, two action steps, and one sentence I can say at the staff meeting without being booed by the copier.”

Expert prompt: “Create a plan that satisfies compliance, respects teacher workload, does not require new funding, can be implemented by Monday, and assumes half the key people are testing.” The chatbot will respond with “Certainly,” which is how you know it has never worked in a school.

The real trick is not getting a perfect answer. It is getting a usable first draft, then applying the ancient human technology known as judgment.

Dear Dr. Clipboard, A teacher wants to ban all AI. Another wants students to build chatbots for every assignment. The board wants “guardrails.” What do I do?

Dear Guardrailed, Begin with principles, not panic. Define acceptable use by task, age, privacy, and learning purpose. Then write guidance short enough that someone might actually read it. Revolutionary, I know.

Dear Dr. Clipboard, A parent says AI will destroy childhood. Another says we are failing to prepare students for the future. They are both copying me on the same thread.

Dear Threaded, Reply once. Use words like balance, discernment, and human judgment. Then archive aggressively.

Dear Dr. Clipboard, The AI tool says it can grade essays instantly. Should we use it?

Dear Instant, Only if you enjoy phrases like “algorithmic bias review,” “FERPA implications,” and “why did my child receive feedback from a toaster?” Use AI to support feedback, not replace professional evaluation.

Dear Dr. Clipboard, I used AI to draft the AI policy. Am I the problem?

Dear Possibly, Not if humans reviewed it. Definitely if the final line says, “As a large language model, I cannot enforce hallway expectations.”