Sunnyvale School District

Special education has never had a shortage of information. Educators navigate student plans, evaluations, service schedules, family communication and compliance requirements while trying to remain fully present for students. Artificial intelligence may help reduce some of that administrative weight. But the most important leadership question is not, “What can AI do?” It is, “Which burdens can AI safely reduce without weakening relationships, professional judgment or student agency?”

In special education, technology is never simply a productivity tool. Its use can affect how a student is understood, what opportunities are offered and whether a family feels included in decisions. Purposeful adoption must begin with human needs, not product features.

Start With the Problem, Not the Platform

Strong implementation begins by listening to the people closest to the work. Ask special educators, related service providers, paraprofessionals, students and families where systems create friction. The answer may be repetitive documentation, inaccessible materials, fragmented communication or difficulty turning data into timely action.

That has been the starting point for the tools I have built in my district. Existing systems reported students receiving special education services, but not those who were evaluated and did not qualify. That missing population can reveal a great deal about the referral process. I built a dashboard combining several data sources so leaders can see both groups and ask better questions.

I also developed a legal reference tool for school psychologists. It answers from state and federal law and relevant case law, with citations to the underlying sources. When the available authority does not answer a question, the tool says so and directs the user to counsel or a district special education administrator. That refusal is an important feature. In a high-stakes setting, an answer that cannot be verified may be worse than no answer.

“Technology should make the system more humane, not merely more efficient.”

Other projects include a privacy-conscious dashboard for examining student growth and inclusion, plus automations for scheduling, flagging messages and organizing employee evaluation notes. The tasks differ, but the purpose is the same: reducing friction so educators and leaders can spend more time in classrooms and with the people they serve.

Build the Conditions for Responsible Use

Access to an AI tool is not an implementation strategy. Leaders must create the conditions for responsible use.

Staff need clear guidance about which tools are approved, what information may be entered, which uses are prohibited and when human review is required. Educators should not have to make institutional privacy and risk decisions on their own. In the tools I have built, student names are not stored. Identifiers connect records while the link between an identifier and a child stays in the district’s system of record. Small groups are also suppressed when reporting them could risk identifying a student. Privacy is part of the design, not a compliance step added later.

Practical AI literacy is just as important. Training should go beyond learning how to write a prompt. Educators need practice checking accuracy, recognizing bias, protecting confidentiality, evaluating accessibility and understanding when a tool has reached the limits of what it can reliably do.

That last point matters because a system can appear to work even when it has failed. I have seen automations report that they were running while producing nothing because the check confirmed only that the process was active, not that the expected work had arrived. Now I ask a simple question before relying on a tool: How will I know if it is wrong? Leaders should ask vendors and internal teams the same question.

Keep Human Judgment at the Center

AI can help organize information, create a first draft, make complex language more accessible or surface a pattern worth examining. It should not independently make high-stakes decisions about a student’s eligibility, services, placement, behavior or potential. Technology can support preparation and analysis. It cannot replace the knowledge of educators, the voice of families or the lived experience of students.

The standard for success should be straightforward. Did the tool improve clarity? Reduce workload? Preserve accuracy? Increase access? Strengthen trust? If not, leaders should be willing to revise the process or stop using the tool.

The promise of AI in special education is not automation for its own sake. It is the possibility of returning time and attention to the work only people can do: building trust, noticing nuance, solving problems together and holding high expectations for every learner.

Technology should make the system more humane, not merely more efficient. That is the purpose leaders must protect as we decide what to build, what to adopt and what not to use.