ShipIn CEO: AI FleetVision Cuts Safety

Osher Perry, founder and CEO of ShipIn Systems, has responded to an article questioning whether AI cameras can measure safety culture. Perry agrees that a camera or a personal protective equipment detection algorithm alone cannot measure culture, competence, or fatigue. He states the camera is merely a sensor.

However, Perry argues that when visual data is combined with other operational information, it moves from simple event detection to pattern understanding. This approach mirrors aviation's long-standing use of routine flight data analysis to identify adverse trends before they become accidents.

From Events to Patterns

The core principle is to measure frequent occurrences to better understand rare, high-impact events. Perry notes that fatigue, workload, and commercial pressure lead to shortcuts. Most shortcuts result in nothing, but some become near misses or contribute to major incidents. AI systems that combine thousands of signals can reveal the high-frequency, lower-impact deviations that precede serious accidents.

A recent joint Maritime Risk Report with insurer NorthStandard found more than half of incidents involving human error also involved systemic factors like fatigue and communication breakdowns. Perry states that one shortcut reveals little, but thousands of observations reveal a pattern.

Data Reveals Accident Risk

Perry addresses concerns about oversimplified metrics, referencing Goodhart's law. He argues this is a case against poor key performance indicators, not against measurement itself. The analysis with NorthStandard examined roughly 600 vessels and 1,600 reported accidents.

The data showed that 29% of vessels accounted for 70% of accidents. Furthermore, vessels with FleetVision scores below 75 experienced an accident rate 5.7 times higher than vessels scoring above 85. Perry emphasizes this indicates recurring operational patterns provide meaningful insight into how a vessel operates, beyond simplistic helmet detection.

Empowering Crews with Continuous Data

Perry contrasts continuous operational data with traditional methods like annual inspections and claims histories, which he calls valuable but lagging snapshots. He asserts that continuous data on key safety parameters fills the gap between these periodic assessments.

He stresses that the technology must empower crews, not police them. "If AI becomes another tool for shore management to police seafarers, we will have failed," Perry writes. The goal is enhancing safety through collaboration between seafarers and shore teams.

Measurable Impact on Operations

Perry provides specific figures on the impact of the FleetVision system. Across hundreds of vessels using the platform for 12 months, operational safety deviations fell from approximately 230 per vessel per month to just eight.

This represents a 96% reduction. Today, FleetVision operates on about 1,300 vessels for 92 shipowners. The technology organizes visual and operational data across bridge, safety, technical, security, and cargo operations to give a comprehensive picture.

The system also reduces administrative work by automatically generating operational logs for crew review. These logs can be used for enhanced training ashore. Perry concludes that safety culture is made of thousands of daily decisions, which a camera cannot measure. But he argues AI can identify patterns showing where risk is building, giving crews information to act earlier.