School-based SLPs face documentation that carries legal weight: IEPs, evaluations on statutory clocks, progress monitoring cycles, Medicaid service logs, and a constant stream of parent communication. LLMs genuinely reduce the friction here — and schools are also where AI drafting has a failure mode with a legal name.

Regulatory Context

• IDEA governs evaluations, eligibility, and IEP requirements — including the timelines: 60 days (federal default; your state may differ) from consent to initial evaluation, annual IEP reviews, triennial reevaluations. AI can’t move these clocks, but it can help you keep documentation from being the bottleneck.
• FERPA protects student education records; no identifiable student data in public AI tools. District-approved tools with appropriate agreements are a different category — know which yours are.
• FAPE means goals must be individualized and meaningful. IEP goals are legal commitments of district resources; precision matters.
• Medicaid billing in schools requires service documentation that matches what was delivered. Generated session notes that drift from what actually happened aren’t just sloppy — they’re audit findings.

The Predetermination Trap

Here’s the schools-specific danger: an IEP developed before the meeting violates the parent’s right to participate. Courts have found predetermination when teams arrived with finished documents and treated the meeting as a formality. An LLM makes it effortless to walk in with beautiful, complete, final-sounding goals — which is exactly the artifact you don’t want in the room. Draft inputs (present levels from your data, goal areas, options to discuss), not outcomes. The polish of AI output works against you here: “DRAFT — for team discussion” should be true, not decorative.

Worked Example: Present Levels From Real Data

“I am a school-based SLP drafting the speech-language section of present levels (PLAAFP) for an IEP meeting. Here are my data: [de-identified assessment scores, classroom observation notes, teacher input, progress monitoring data]. Organize into a present-levels narrative that states what the student CAN do, how the disability affects access to the general education curriculum, and what the data shows. Use only the data I provided — do not add strengths, needs, or classroom impacts I did not report. Label it DRAFT. Keep my clinical voice.”

What to check: The curriculum-impact statement is where models invent — watch for plausible classroom effects (“struggles to follow multi-step directions during instruction”) that came from the model’s training data, not your observation. Every sentence should trace to a data point you provided, because every sentence may be read aloud at a due process hearing.

Red Flags

• Entering identifiable student information into public models
• Walking into an IEP meeting with AI-polished, final-sounding documents — the predetermination trap
• Using AI for eligibility or diagnostic decisions
• Medicaid service logs that describe generated content rather than delivered services
• Polished wording hiding non-measurable goals
• Letting fluent language disguise weak reasoning

Key Workflows

• Eval season: assessment report structuring, score-summary narratives — see the eval report workflow
• IEP prep: present-level drafting from data, goal options for team discussion — IEP meeting prep
• Progress monitoring: note organization, data summaries in batches — progress monitoring workflow
• Parent communication: email drafts, jargon-free summaries

Related

• Prompts: IEP Goal Strengthener · Progress Note Organizer
• Goals: the goal bank — school-setting goals with four-questions structure and evidence bases
• Policy: School District AI Policy template

See it in practice: Case study — eval season, 8 referrals in 3 weeks.