---
title: AI & student data
category: procurement
order: 6
lastUpdated: 2026-06-08
featureAvailability: GA
roles: [district-admin, admin]
relatedArticles: [ai/ferpa-ai-data-handling, ai/ai-providers, procurement/ferpa-overview, procurement/subprocessors]
---

## At a glance

- **Who uses it:** District IT, legal, and DPO reviewers evaluating AmplifyED's AI features for student data exposure.
- **Purpose:** Plain-language procurement summary of how AI handles student data — and how your district can opt out.

## Summary

AmplifyED uses AI to draft notes, generate case summaries, and propose action plans. AI is **opt-in feature surfacing** — every AI output is a draft until a human accepts it, and every core workflow works without AI.

## Data flow

1. A facilitator requests an AI draft or summary inside a case.
2. The relevant case text is sent to AmplifyED's managed AI gateway over TLS.
3. The gateway calls the underlying model (Google Gemini 3 Flash Preview) in a US region.
4. The model returns a draft. The draft sits in the UI until the facilitator accepts it.
5. If accepted, the draft becomes a normal note (visibility rules apply). If discarded, no AI output is retained.

## What we do not do

- **No model training.** Customer student data is never used to train the underlying AI model. This is contractually enforced with the AI provider.
- **No long-term storage at the model.** Inputs and outputs are not retained by the model provider beyond the transient processing window required to return a response.
- **No external sharing.** AI requests are not visible to other customers, to AmplifyED platform staff (outside of redacted, aggregated telemetry), or to third parties.
- **No auto-application to records.** AI never writes to the case record without a human accept.

## Opting out

A district can disable AI org-wide by contacting your account manager. The platform retains every non-AI capability. If your district opts in later, no historical data is reprocessed — AI begins applying only to new requests.

## Audit posture

- The acceptance of an AI draft is audit-logged (`note_added` with metadata indicating AI origin).
- Discarded drafts are not retained.
- Aggregated, non-PII usage telemetry powers the monthly usage display on Administration → Billing.

## Related articles

- [AI & FERPA data handling](/help/ai/ferpa-ai-data-handling)
- [AI providers and models](/help/ai/ai-providers)
- [FERPA overview](/help/procurement/ferpa-overview)
- [Subprocessors](/help/procurement/subprocessors)

## FAQ

**Can we limit AI to certain roles?**
AI today is gated by plan, not by role. We are scoping role-level controls for a future release; contact your account manager to influence priority.

**Is there a way to test AI on synthetic data first?**
Yes. We recommend running a 2-week pilot in a non-production org seeded with synthetic students before enabling AI on the live tenant.