Early intervention (birth to three, under Part C of IDEA) inverts the usual LLM problem. In most settings, the risk is that casual notes get formalized badly. In EI, formalizing is itself the failure mode. IFSP outcomes are supposed to be written in the family’s own words, about the family’s own routines — “Maya will let us know what she wants at dinner so mealtimes stop ending in tears,” not “child will increase expressive communication to 80% accuracy.” A language model’s strongest instinct is to convert plain language into professional language. Here, that instinct works directly against compliant, family-centered practice.

The Practice Realities

• Outcomes belong to the family. Part C requires IFSP outcomes based on family priorities, in functional, routine-based terms. If the output sounds like a school IEP goal, it’s wrong — not stylistically, structurally.
• You coach; you don’t drill. The evidence-based EI model builds caregiver capacity in natural environments. Session notes should document what the caregiver practiced, what strategy they’re carrying into the week, and how the routine responded — not clinician-directed trial data.
• Timelines are real: the 45-day clock from referral to IFSP, six-month reviews, and transition planning before age three (to Part B preschool services or out). Documentation gaps surface at every one of these checkpoints.
• Developmental variation at this age is enormous. LLM-generated “norms” flatten it. A model will happily pathologize a late talker or reassure about a child who needs referral — with equal fluency and zero accountability.
• No coercion is a floor, not a preference. Feeding and communication outcomes for infants and toddlers must respect the child’s cues and the family’s agency. Watch generated activity ideas for compliance-flavored framing (“have the child…,” “require,” “withhold until”).

Worked Example: A Coaching-Model Session Note

“I am an early intervention SLP using a coaching model under IDEA Part C. Here are my notes from a home visit: [de-identified notes — the routine you targeted, what the caregiver tried, the child’s response, the plan the caregiver chose]. Organize into a session note with: routine and setting, caregiver strategy practiced, child’s response, caregiver reflection, and joint plan. Keep the family’s own words for their routines and priorities — do not convert them into clinical terminology. Do not add developmental interpretations I did not provide.”

What to check: Did “snack time keeps turning into a fight” survive as itself, or did it come back as “child demonstrates mealtime behavioral dysregulation”? The family’s phrasing is the compliant version. Also check that the plan reads as the caregiver’s plan — if the model has recast coaching as clinician-delivered treatment, the note now misrepresents the service model you’re billing under.

Red Flags Specific to This Setting

• IFSP outcomes that sound clinical instead of familial — the LLM “improving” language into non-compliance
• Over-pathologizing typical variation based on generated milestone claims — verify ages against a real source, never a model’s memory
• Activity suggestions that ignore the family’s actual routines, resources, or culture
• Substituting generated handouts for the coaching conversation the model of service requires

Related

• Workflow: Home Program Creation — the family-context reasoning applies directly
• Prompt: Parent-Friendly Summary
• Goals: the goal bank’s pediatric feeding goals model the no-coercion, family-agency conditions EI requires