Districts should start with AI language, not tools

start AI – A lesson from two years of AI literacy work across K-12 schools in São Paulo suggests American districts are chasing platforms too early. The shift that accelerated teacher uptake came from addressing one fear head-on in faculty meetings—teachers worried they

When a senior mathematics teacher with 22 years in the classroom raised her hand at the end of a staff meeting in early January, she didn’t ask about AI tools. She asked about what it would cost her socially.

“What if I look stupid in front of my students?” she said.

The room went quiet. Nobody had said it out loud before. but every teacher present carried some version of the same worry for months—American districts trying to build a shared AI structure. the speaker implied without preaching. were spending too much time on tooling and too little on the question that actually determines whether teachers will use what they learn.

Over two years. around 50 K-12 colleagues across three international schools in São Paulo worked on AI literacy. guided by the same practical question: what changes behavior inside classrooms. and what doesn’t. In that experience, the dominant barrier to adoption was not technophobia, not generational mismatch, and not fear of replacement.

The problem was more pointed. Experienced teachers feared being seen by their students as the last person in the room to understand a tool students were already using. Naming that fear in a faculty meeting—and giving teachers explicit institutional permission to learn alongside their students—accelerated uptake sharply at around the eight-month mark.

The lesson for American district leaders was blunt: the language shift can be done before any procurement decision, at zero marginal cost.

That gets to a second design choice, one that determines whether learning lasts or stays as workshop energy.

Whole-school AI professional development days, in the São Paulo cohort, produced low durable change. Teachers attended, took notes, and returned to classrooms with little visible behavioral shift six weeks later. Self-directed learning produced uneven change concentrated among already-willing teachers, widening internal gaps instead of closing them.

The clearest behavioral signal came from department-level structured engagement. In groups of four to eight teachers, the program ran across four sessions over six weeks. Each meeting included one practice task to complete in a real lesson between gatherings. followed by one shared observation at the end. The template was specific: 45 minutes per session. one specific pedagogical question per session rather than one tool per session. and a final shared observation written up in two paragraphs and circulated to the rest of the faculty.

The sequencing also mattered. The work did not begin with departments that initially resisted. Instead, it started with two willing departments, published a short internal write-up of what changed, and only then allowed resistant departments to approach when they were ready.

A third decision—how a district frames AI use itself—was where policy language met classroom reality.

The most damaging framing in current U.S. K-12 policy, as the São Paulo experience describes it, is the binary question: did the student use AI or not. That either/or model cannot hold up inside real classrooms.

A mathematics student using AI to check work before submission is doing something different from a student using AI to bypass the work entirely. A history student using AI to summarize a primary source is doing something different from a student using AI to substitute one. Treating both as the same category turns instruction into policing.

The cohort that showed the strongest uptake treated AI use as a competence within a discipline. built around observable criteria specific to each subject. Drafting those statements. in practice. took less time than most district leaders expect: one paragraph per discipline with three to five observable criteria. written by the head of department and signed off by the principal in around 90 minutes.

The language standard was deliberately student-centered. The statement should be in language a 14-year-old can read, not in the language a lawyer drafted. When students can read the criteria, they self-regulate against them. When they cannot, they cheat against them.

Even when districts get buy-in, the order of operations can still derail adoption.

Most districts begin with tooling: evaluate three platforms. pick one. roll it out. then wait—only to discover uneven teacher uptake six months later. In the São Paulo cohort, the sequence worked in the opposite direction. Start with the language leaders use about AI in faculty meetings. Then move to the structure of department-level engagement. Next come discipline-specific competence statements. Only after those steps should a platform be chosen.

When the platform is chosen. the experience stresses. it must be chosen with the heads of department who will actually use it—not with an IT committee deciding in their absence. In that approach, language, structure, and competence statements determine whether a district sees a return on any platform. If the platform arrives first and the other three pieces are wrong. the result is predictable: a budget line. but not the behavior change.

One guiding paragraph connects it all: when teacher fear is named, when engagement is structured where work actually happens, and when AI use is defined in discipline-specific, student-readable competence criteria, the platform decision stops being the center of gravity.

For a district leader staring at the calendar, the message is immediate and specific. In the next faculty meeting, change the wording about AI from “we will permit it under the following conditions” to “we will learn it alongside our students, and here is what that looks like.”

Then propose to two department heads a four-session structured engagement with measurement at the end. and offer to attend the first session personally. Ask one head of department to draft a single discipline-specific AI competence statement in plain language as a template for the rest of the faculty.

None of the approach requires money the district does not already have. What it requires is the leader changing the language they use in faculty meetings. being honest about which budget lines have produced behavioral change and which have not. and accepting that AI literacy in a district is not a procurement project.

It is a language project. A structure project. A competence project. Start there—tomorrow—before the next budget cycle arrives.

AI literacy K-12 education teacher uptake faculty meetings department-level engagement AI competence statements school policy São Paulo international schools education technology