# Get Simpl AI healthcare suite — synthetic diabetes walkthrough
This public browser simulation uses the existing wearable/FHIR normalization and deterministic summary modules. It does not connect production accounts, access a database, authenticate clinicians or call an external model.

1. Connect the synthetic watch. Point out the missing 10 October sample and low-quality sleep data.
2. Authorize sharing explicitly, then transfer. Repeat transfer to demonstrate idempotency.
3. Review the timeline: consumer wearable signals, imported clinical HbA1c results (7.8% then 7.3%), diabetes history and encounter, patient-reported metformin and penicillin rash, and synthetic report metadata.
4. Review sources. Generate the source-linked Clinical draft. Inspect engine version and source IDs.
5. Explain stale laboratory data, quality limits, and the distinction between numeric change and clinical significance.
6. Record the simulated reviewer assessment. Download the reviewed evidence JSON including session audit.
7. Revoke sharing; generation and export become unavailable. Downloaded copies cannot be recalled.

All people and records are fictional. Review is simulated, not a real clinician signature. Browser audit is not tamper-proof. No diagnosis, treatment recommendation, prediction, medical validation, regulatory approval or hospital partnership is implied. Production integration requires authenticated identities, verified tenant/patient bindings, source-specific grants, server-enforced access, durable review and audit. Existing institutional API supports those boundaries; this demo is not proof that independent product accounts are connected.

## Proposed institutional pilot
Start with synthetic records and institution security/legal review. Appoint a clinical sponsor and data controller. Agree intended use, residency, retention, incident response, source mappings and model governance before any real-patient phase. Then use a limited, consented retrospective cohort under local governance, with clinicians independently assessing source accuracy, missing-data handling, summary faithfulness and review burden. Predefine acceptance criteria and stop rules. Keep all output advisory, with no automated care actions. A prospective phase requires separate institutional approval and evidence review.