One CCTV Alert: What Does COSMOS Do Next?

Hong Kong’s construction-site smoking ban is now in force. For contractors, the challenge is not limited to detecting someone smoking. It is maintaining effective no-smoking controls across a site that changes every day.

Work fronts move. Scaffold levels open and close. Temporary access routes, plant zones and material storage areas shift as construction progresses. With multiple subcontractors and trades working simultaneously, manual patrols cannot provide continuous visibility across every corner.

AI CCTV can help surface a potential smoking event. But an alert only captures one moment.

What happens next determines whether that moment contributes to site-wide prevention.

Why smoking detection alone is not enough

The new requirements place clear responsibilities on principal contractors, subcontractors and people on construction sites. Contractors and subcontractors must take reasonable steps to prevent smoking and the use of naked lights, helping reduce fire risk and protect workers and the public. Read the HKSAR Government announcement.

A CCTV alert can indicate that potential smoking behaviour may have occurred. By itself, however, it cannot establish:

Whether the event has been reviewed by an authorised person
Whether it is a confirmed event or a false positive
Whether similar events have occurred in the same area
Whether no-smoking signs and briefings are current
Whether follow-up action has been assigned
Whether the action has been completed and recorded

Without a connected workflow, alerts remain isolated notifications. Teams may respond to individual incidents without seeing recurring locations, time patterns or gaps in existing site controls.

Detection is the starting point. Prevention depends on the action that follows.

What COSMOS does after an alert

COSMOS is Varadise’s connected operating platform for smart construction. It brings CCTV and sensor signals, project records, site locations and operational workflows into one management environment.

Athena AI works as the platform’s intelligence layer, helping teams surface relevant information, review control records and coordinate follow-up. Human reviewers remain responsible for confirming events and making final decisions.

The smoking-management workflow can be structured around five stages.

1. Surface the potential event

When AI CCTV identifies potential smoking behaviour, COSMOS creates an event record containing the available operational context, such as:

Time of detection
Camera source
Site location or work zone
Supporting footage or images
Related sensor signals, where available

The system should identify the person or event for review—not treat an individual cigarette or an AI prediction as automatic proof of a confirmed violation.

2. Support human review

An authorised team member reviews the available evidence and classifies the event.

The reviewer may confirm that follow-up is required, mark the footage as inconclusive or close a false positive with an appropriate reason. COSMOS records who reviewed the event, when the review occurred and what decision was made.

This creates a clear separation between machine detection and human judgement.

3. Check the surrounding site controls

A reviewed event should prompt a wider question: was it an isolated act, or could it indicate that existing site measures require attention?

Using project-defined requirements and available records, Athena AI can assist the team in checking whether supporting evidence is available for:

No-smoking signs in the relevant area
Site inductions for workers
Toolbox talks and safety briefings
Patrol and inspection records
Previous events in the same location
Earlier corrective actions and close-out records

Athena AI can surface missing, outdated or inconsistent information for human review. It does not replace the professional judgement or legal responsibilities of the relevant duty holders.

4. Identify recurring smoking-risk patterns

One event shows what happened at a specific moment. Multiple reviewed events can reveal a pattern.

By connecting events to site zones and time periods, COSMOS helps teams examine whether potential smoking behaviour is:

Recurring in the same location
Concentrated around secluded or temporary work areas
Appearing during particular work or break periods
Increasing after a work front or access route changes
Continuing despite previous follow-up

This information enables teams to adjust their controls based on current site conditions. Actions may include repositioning signs, changing patrol routes, conducting targeted briefings or reviewing supervision arrangements.

5. Assign and close follow-up actions

COSMOS connects reviewed events and identified control gaps to structured actions.

The team can record:

What action is required
Who is responsible
The target completion date
Supporting photographs or documents
Completion status
Final review and close-out

Instead of relying on scattered emails, chat messages and spreadsheets, the complete workflow remains connected:

Alert → Review → Control check → Action → Evidence → Close-out

Project leaders can see which events await review, where risks are recurring and which actions remain outstanding.

From one alert to site-wide prevention

Consider a potential smoking event identified near a temporary material storage area.

The event is surfaced by AI CCTV and mapped to the relevant work zone. A safety team member reviews the footage and determines that follow-up is required.

Athena AI then assists in checking whether the area has current signage, briefing records and recent patrol evidence. The team identifies that the site layout changed recently and the original sign is no longer clearly visible.

An action is assigned to reposition the sign, brief the relevant work team and update the patrol route. Completion evidence is uploaded and reviewed before the action is closed.

If further events occur in the same area, COSMOS helps the team recognise the recurring pattern and review whether additional measures are needed.

The value is not the alert alone. It is the connected process that turns the alert into operational action and learning.

Move beyond one-point detection

AI CCTV helps construction teams see potential smoking events that manual patrols may miss. Sensors and field observations provide additional signals.

COSMOS connects these signals with human review, Athena AI-assisted control checks, risk trends and traceable follow-up.

This helps teams move from reacting to individual CCTV alerts toward a continuous, site-wide smoking-prevention process.