Use case · Continuous AI assurance

Watch an assurance agent audit in real time.

Traditional assurance is a point-in-time audit — a snapshot that's stale the day the report lands. A digital assurance agent watches the evidence continuously. Below: a simulated live assessment of a fictitious infrastructure operator's AI video-analytics system against the NSW AI Assessment Framework and ISO 55001.

System under assurance

AI video analytics — restricted-area intrusion detection, 13 sites

Frameworks applied

NSW AIAF (2026) · ISO/IEC 42001 · ISO 55001

Evidence sources

Model registry · maintenance records · incident log · governance docs

assurance-agent · idle

Assurance
confidence

—

Assessment complete — the picture a quarterly audit would have missed.

Evidence items checked

Conformant

Flags raised

Blocking gap

The blocking gap: the detection model was retrained in May, but no AIAF re-assessment followed — under the mandate, a material model change triggers reassessment before continued operation. A quarterly audit cycle would have surfaced this in September. The agent surfaced it in 90 seconds, drafted the corrective action, and assigned it. That's the difference between assurance as a report and assurance as a live control.

How it works under the hood.

Ingest evidence continuously

The agent connects to where evidence already lives — document stores, asset systems, model registries, incident logs. No new data entry.

Assess against the frameworks

Every item is checked against clause-level requirements — NSW AIAF, ISO/IEC 42001, ISO 55001 — with the reasoning shown, not hidden.

Flag, draft, track

Gaps become drafted corrective actions with owners and due dates. The assurance picture stays live between audits — not frozen inside last quarter's report.

This demonstration uses simulated evidence. In an engagement, the same assessment runs against your real documents and systems — confidentially, inside your environment.

Book a confidential conversation →