When loss-prevention leaders talk about productivity wins, the metric you hear most often is 'CCTV review time'. The benchmark goes: we used to spend 45 minutes scrubbing through footage to find an incident; now we spend two. The implied story is that the AI made the search faster.
That story is roughly right but operationally incomplete. The interesting question isn't how long a manager takes to find the clip, it's how long from the moment something happens until a forward-ready evidence pack is in the right inbox.
What 'forward-ready' actually requires
A police force won't act on a clip alone. They want context: which camera, what time, who was involved, what happened before and after, and, critically, whether this person has been involved in prior events you can evidence.
A forward-ready pack therefore has three components:
- Multi-camera chain, the subject tracked from detection through to exit, with all camera IDs and timestamps logged.
- Append-only audit log, every status change since the event was created, with the user, role, and timestamp.
- Prior-event linkage, if the face matches one we've seen before, the prior events are attached automatically with their own evidence.