How we use AI

Structured IEP Workspace uses large language models to support, not replace, the professional judgment of IEP teams. This page explains where AI is invoked, what it does, and what humans must review.

Where AI is used

• Document extraction: pulling structured fields from prior IEPs and evaluations.
• PII detection: flagging values for redaction. Final decisions stay with the user.
• PLAAFP drafting: drafting narrative based on user-entered data and accepted extracted data.
• Strength and need suggestions: proposing items the user can accept, edit, or reject.
• Goal scaffolding: drafting measurable annual goals from needs and baselines.
• Validation explanations: explaining findings and suggesting rewrites.

What AI is not used for

• Eligibility determination.
• Diagnosis or classification of disability category.
• Autonomous IEP generation. No section is finalized without explicit user action.
• Compliance reporting or submission to a district system of record.

Models

AI calls are routed through the Lovable AI Gateway. Current default models include Google Gemini 3 Flash Preview for drafting and validation assistance. Models may be updated; user data is not used to train third-party models.

Human review

Every export package includes a reviewer disclaimer noting that the draft was created with AI assistance and must be reviewed by a qualified IEP team member before use. Validation findings and citations are surfaced so reviewers can verify each section against its source.