technology
From First-Person Video to Procedural Intelligence
We combine VLM and state machine technology to analyze worker actions, assess procedures, and document the results as evidence.
Real-time Analysis
≤ 3 sec
Built to Scale
1,000+ concurrent users supported
What Makes Procedure AI Different
More Than One Moment
We analyze multiple moments in order to understand the complete procedure.
More Than a Fixed Sequence
Understand conditions, branches, and repeated steps.
More Than One Domain
Adapt to different procedures across different domains.
First-Person (Nexicle) vs Third-Person (CCTV)
First-Person (Nexicle)
What the worker sees
Capture exactly what matters.
Fewer blind spots
No important details are missed.
Understand What Happened
See the actions and steps that led to the result.
Third-Person (CCTV)
Limited viewpoint
You see only what the camera sees.
Blind spots
Important actions are often hidden.
Missed details
Hard to know exactly what happened.
How Procedure AI Works
1. First-Person Input
Video / Audio
2. VLM
Understand actions, objects, and context in natural language.
3. State Machine
Understand procedure logic: sequence, conditions, loops, parallel steps, time checks.
4. AI Evaluation
Action Recognition, Sequence Check, Timing, Compliance
5. Evidence & Report
Generate evidence-based evaluation and feedback with reasoning.
Continuous Improvement
The more real procedures we analyze, the smarter Procedure AI gets.
- ✓ Collect real-world execution data
- ✓ Refine SOP logic and evaluation rules
- ✓ Expand procedure scenarios
- ✓ Improve accuracy and adaptability
Performance & Scalability
≤ 3 sec
Per-step evaluation time
1,000+
Concurrent users supported within a single organization