Cases and video intake
Each case holds multiple videos, with unit, camera and retention set at creation. Source folders can be read in place, without copying.
AMAS AI Video Governance PlatformStop scrubbing through hours of footage.Start with what matters.
AI ranks what to review; your team makes the call. "Not yet reviewed" is never confused with "reviewed, nothing found" — and every view, review and export is logged.
Works on recorded video, not live alerts. Every final call is made by a person against the original footage.
- First passAI ranks; it does not judge
- Every flag reviewedA person agrees or disagrees
- Two ledgersUnreviewed ≠ all clear
- Tamper-evidentViews, reviews, exports logged
01 — The problem
Footage keeps piling up. People can only spot-check.
For public agencies and their contractors who must review large volumes of recorded video — and want AI to share the load without taking over accountability.
- 01
More footage than people
Cameras record all day; reviewers can only sample a fraction of it.
- 02
AI without handing over judgement
Public bodies need a clear line of accountability: AI can rank, but people must write the conclusion.
- 03
Results that can't withstand scrutiny
"Zero findings" cannot tell "all clear" apart from "never checked".
02 — Capabilities
From intake to delivery, a person owns every step.
The platform does the first pass only. It flags time ranges worth a look — never a verdict, and never grounds for a penalty on its own.
Review workspace
Video, a density timeline and each finding side by side. Every flag needs a person to agree or disagree, and edits never overwrite the original reading.
Delivery with a quality gate
What you preview is what you deliver. Formal delivery is blocked while any item is unreviewed; exports carry a watermark, disclaimer and audit record.
Childcare camera-setup check
Checks camera footage against childcare CCTV regulations item by item, and clearly marks what cannot be measured or is not the system's call.
Video privacy governance
Per-case retention with a 30-day floor, legal hold, cell-level permission changes with a required reason, and every view logged.
Built-in knowledge base
Regulations, policies and reviewer experience pass a staged intake with a mandatory human gate. Review conclusions can become reusable knowledge, and search shows which sources were checked.
03 — How it works
AI sorts first. People decide last.
- 01
Create a case
Add videos or scan a source folder.
- 02
AI first pass
Sampled frames analysed, events graded by duration, time ranges flagged.
- 03
Human review
Reviewers agree or disagree with each flag and record why.
- 04
Deliver and retain
Export once the quality gate passes; conclusions become knowledge for the next reviewer.
04 — Governance & trust
Accountability stays with people. The record stays in the system.
Video is highly sensitive personal data. The platform records who saw what and who decided what, so the work can withstand scrutiny later.
Flags, not verdicts
The output is a list of ranges to review; whether anything happened is decided by a person against the original footage.
Tamper-evident audit trail
Views, reviews, permission changes and exports are recorded, and later alteration can be detected.
Retention and legal hold
Retention is set for each case at creation, with legal hold when needed.
Deployment options
Inference can run on local GPUs or domestic cloud compute. In on-site engagements we keep no copy of the video, and extracted frames are deleted after processing.
05 — Where it stands
Current statusPilot projects
Delivered as projects, one application at a time.
The platform is deployed in production and can be demonstrated, and it has completed first-phase on-site review in one public-sector project. Only the childcare review application is live today; the others are labelled below.
There is no self-service trial. Every engagement starts with an assessment of your video sources, storage and review staffing.
- DonePhase 1: on-site reviewWe bring the equipment, read the source video and keep no copy.
- NowPilot projects · demo available
- PlannedPhase 2: appliance deliveryPlatform and inference installed in your server room; you decide what is kept and deleted.
Review applications
Shown with the same three statuses the platform itself displays.
- LiveChildcare video reviewFlags time ranges in childcare-centre recordings and prompts reviewers to check the original footage.
- Capability proven · not on platformPPE workplace-safety reviewProduct line on hold.
- PlannedLabour inspection footage
- PlannedPersonal safety
- PlannedDrone patrol
- PlannedTraffic footage
06 — The AMAS family
Same governance principles, different front lines.
Each product runs on its own, and each keeps the same rules: clear roles, traceable records, and AI that assists but never decides.
Available
AMAS Enterprise AI Knowledge Governance Platform
Every answer cited. Every access governed.
Learn moreEarly access
AMAS Corewatch Security Governance Platform
From "are we affected?" to audit-ready, in one workflow
Learn moreDemo available
AMAS Enterprise HR Governance Platform
Payroll calculated to Taiwan's rules, with maker-checker controls
Learn moreNext step
Start from your review workflow.
Delivered as a project: we start from your site and footage sources, then assess which triage application and review flow fit.