The AI Intelligence Suite
Voice narrative, population risk, staffing-shortfall prediction and regulatory returns — every engine grounded in live facility data, streaming in real time, and built for the operational reality of correctional management.
Not AI bolted on. AI grounded in your data.
Every engine shares the same foundation — designed for the security, reliability and auditability a correctional environment demands.
Live grounding
Wired directly to facility data — inmate records, sick calls, medication logs and rosters. No cached or pre-generated reports.
Real-time streaming
Progressive token streaming over Server-Sent Events. Staff see output being built with zero queue delay.
No data invention
A strict zero-hallucination guardrail. Missing figures return an explicit “verification required” placeholder — never a guess.
Audit defensible
Every calculation, prompt, output and reviewer modification is logged for a complete, defensible trail.
At a glance
All four engines, their core capability, benefit and regulatory tie.
| Feature | Core capability | Primary benefit | Regulatory tie |
|---|---|---|---|
| AI Voice-Note Studio | Voice/text → structured incident narrative | Faster, compliant incident documentation | PREA reporting; audit-trail integrity |
| AI Risk Radar | Population-wide risk triage from live signals | Early identification of at-risk individuals | NCCHC / PREA duty of care |
| AI Staffing Radar | Shift-by-shift coverage shortfall prediction | Prevent unsafe staffing before it happens | ACA post coverage; PREA ratios |
| AI Oversight Returns | Auto-generated regulatory return drafts | Days of report-writing reduced to seconds | ACA / PREA / CJIS / NCCHC reporting |
Every AI feature — all in one place.
Select an engine to see what it does, how it works, its capabilities and the standards it maps to.
AI Voice-Note Studio
From a raw account to a submission-ready incident narrative — in seconds.
Officers dictate or type a rough, plain-language account of what happened. The AI converts it into a professionally structured, legally sound incident narrative ready for submission — turning one of the highest-risk administrative tasks in a facility into a guided, consistent workflow.
Step-by-step workflow
Key capabilities
- Hands-free speech dictation (browser-native Web Speech API)
- Progressive token streaming — no batch waiting windows
- PREA misconduct screening on generated tokens
- Automated severity scoring and incident triage
- Confidence index and decision-support disclaimer on every narrative
- Live input guidance for essential evidentiary detail
Compliance connection
Raw account in. Compliant narrative out.
A real Voice-Note Studio transformation.
“Davis started hitting the wall in his cell around 2200, saying he wanted to hurt himself. Medical was called, Mitchell went in with me, we talked him down. He’s on the watch list now. He had a razor under his mattress.”
At approximately 2200 hours, Inmate Davis (Block D, Cell 12) was observed striking the cell wall repeatedly and making statements indicative of self-harm intent. Officers Mitchell and [Reporting Officer] responded; medical staff were notified and attended. The inmate was de-escalated without physical force. A welfare search recovered a razor blade concealed beneath the mattress — seized and logged. Davis was placed on enhanced observation pending mental-health review.Severity: HIGH · PREA flag: Not applicable · Action: MH referral within 24h; welfare rounds every 30 min.
How the four engines work together
An incident is documented into a compliant narrative and recorded.
That signal re-scores the population and updates the risk watchlist.
Coverage needs for higher-risk populations feed the staffing forecast.
Every data point rolls up into audit-ready regulatory returns.
Mapped to the standards that matter
ACA, PREA, NCCHC and CJIS — each engine as a primary or supporting control.
ACA Standards
Post coverage, grievance timelines and incident-logging timeliness.
PREA (28 CFR 115)
Vulnerability screening, staffing ratios and annual public reporting.
NCCHC Standards
Suicide prevention, medication adherence and sick-call access.
CJIS Compliance
Encryption, role-based access control and immutable audit trails.
Security, reliability and auditability
Real-time data grounding
Every output is generated from live data pulled at the moment of the request — including incidents from the last hour and medications refused this morning.
Streaming output
All four engines stream progressively via Server-Sent Events, so staff get information quickly instead of waiting for a batch result.
No data invention
The AI never fabricates. Missing figures return “DATA NOT AVAILABLE — manual entry required” rather than a plausible-sounding guess.
Confidence scoring
Every output carries a 0–100% confidence score reflecting how complete and consistent the underlying data was.
Session security
All AI endpoints require an authenticated session. No output is generated without a valid login, and all access is logged.
Human oversight
Every output includes a decision-support disclaimer and a recommendation for authorised-officer review before use.
“The AI provides the draft; the professional makes the decision.”
All AI outputs are clearly labelled as decision-support tools — incident narratives, risk assessments, staffing projections and regulatory returns are designed to be reviewed and approved by an authorised officer before operational use or regulatory submission.