ROI Case File No.318 | 'Asia Marine's Sleepless Sea'

Chapter 1: The Wall of Night—The Cost of 24-Hour Operations

The week after Aqua's LEAN case was resolved, a consultation arrived from Chiba regarding vessel operations at a marine construction company. Case File 318 of Volume 26, "The Pursuit of Reproducibility," tells the story of overcoming labor shortages and fatigue risks through AI-human collaboration.

"Detective, our vessels operate on a 24-hour system. Undersea cable laying, offshore wind power construction... tight project schedules mean we can't stop even at night. However, labor shortages and fatigue risks during night navigation are serious. If the captain naps, monitoring becomes thin. Accident risk increases."

Tatsuya Umino, Operations Manager of Asia Marine Tech Co., born in Chiba, visited 221B Baker Street with exhaustion written across his face. In his hands were vessel operation records and, in stark contrast, an accident report marked "judgment delay due to fatigue."

"We conduct marine construction based in Chiba. Undersea cable laying vessels, crane ships, work vessels... we operate 13 vessels. Projects run 24 hours. Vessels can't stop. But humans can't work 24 hours."

Asia Marine's Operational Challenges: - Established: 2005 (marine construction) - Annual Revenue: 4.8 billion yen - Number of Vessels: 13 - Crew: Average 8 per vessel - Operation System: 24 hours (3-shift rotation) - Night Navigation Problems: Extended monitoring duties, judgment delays due to fatigue - Near-miss Incidents Past Year: 42 cases (32 at night, 76%)

Umino's voice carried deep crisis awareness.

"The problem is that night monitoring duties become extended. Captains work 8-hour shifts, but night visibility is poor and weather conditions change easily. Radar, GPS, weather data... we must continuously monitor much information. But human concentration has limits. Late night at 2 AM, 3 AM... fatigue peaks."

Typical Night Navigation Situation:

One night, 2:00 AM:

Captain A (52 years old, 28 years experience): - Work hours: 10 PM - 6 AM next day (8 hours) - Current: 2 AM (4 hours into shift)

Bridge (wheelhouse): - Captain A on watch - First mate napping (2-hour rotation) - Monitoring radar screen - Checking GPS route - Checking weather information - Adjusting engine output

2:15 AM: Small vessel appears on radar

Captain A: "2km ahead, small vessel... probably a fishing boat. Checking course... distance not closing. Should be fine."

2:30 AM: Weather information shows wind speed increase forecast

Captain A: "Forecast shows wind speed rising from 3m/s to 7m/s... still okay, but attention needed."

2:45 AM: Captain A yawns

"Sleepy... I've been watching screens for over 4 hours. My eyes are tired..."

3:00 AM: Another vessel shadow on radar

Captain A: "Another vessel... distance is... 4km... fine, far away."

However, this judgment was wrong. The other vessel was approaching at high speed. Distance wasn't 4km but 2km. He misread the radar display scale.

3:05 AM: First mate returns from nap

First Mate: "Captain, the vessel ahead is approaching! Distance 1km!"

Captain A: "What! Really... immediate course change!"

Emergency evasive action

Result: Collision avoided, but near-miss report submitted.

Captain A's testimony: "Due to fatigue, I misread the radar display scale. Late-night extended monitoring is truly painful..."

Chapter 2: The Reality of Human Limits—Areas AI Should Supplement

"Umino-san, what countermeasures are you considering to improve the night operation system?"

To my question, Umino answered.

"We're considering increasing crew. Currently 8 per vessel, but increasing to 10 would add margin to the rotation system. However, securing personnel is difficult due to crew shortages. And personnel costs increase."

Current Approach (Personnel Reinforcement Type): - Countermeasure: Increase crew - Problem: Crew shortage, personnel cost increase - Limitation: Human concentration is finite

I explained the importance of AI-human collaboration.

"Just increasing people won't solve it. If we support the OODA loop with AI, humans can concentrate on judgment. AI performs observation, AI presents situational assessment, and the captain makes decisions. This collaboration makes 24-hour operations sustainable."

⬜️ ChatGPT | Catalyst of Conception

"Don't increase people. Combine AI and humans with OODA. Observation by AI, judgment by humans."

🟧 Claude | Alchemist of Narrative

"Responsiveness means thinking calmly amid chaos. AI eliminates noise, captains handle only decisions."

🟦 Gemini | Compass of Reason

"OODA is adaptation technology. Accelerate Observe-Orient-Decide-Act and follow environmental changes."

The three members began their analysis. Gemini unfolded the "OODA × AI Support Framework" on the whiteboard.

OODA Loop AI Support Design: 1. Observe: Sensor arrays automatically acquire weather and route data 2. Orient: AI presents risk indices 3. Decide: Captain uses both AI support and human judgment 4. Act: Adjust route according to situation

"Umino-san, let's redesign Asia Marine's night operations with OODA and AI."

Chapter 3: The Eye of Observation—AI Watches the Sea 24 Hours

Phase 1: Observe - Sensor Array Setup (3 months)

First, we added various sensors to vessels and built a system for automatic data collection.

Existing Sensors: - Radar (detecting other vessel positions) - GPS (own vessel position) - Weather sensors (wind speed, temperature, pressure) - Engine monitoring system

Added Sensors: - AIS (Automatic Identification System): Acquire other vessels' course and speed in real-time - Marine weather data API: Wave height, ocean current, visibility forecast data - Hull sensors: Detect shaking, inclination - Fatigue detection camera: Analyze captain's eye movements, blink frequency

Investment: - Sensor addition: 2.8 million yen per vessel × 13 vessels = 36.4 million yen - Data integration system: 12 million yen - Total: 48.4 million yen

After 3 months: Sensor array installation completed on all 13 vessels

Phase 2: Orient - AI Risk Index Calculation (4 months)

Next, we analyzed collected data with AI and calculated risk indices.

AI Model Design:

Input Data: - Own vessel position, speed, course - Surrounding vessel positions, speeds, courses (AIS data) - Weather conditions (wind speed, wave height, visibility) - Captain fatigue level (camera analysis)

Output: - Comprehensive risk index (0-100) - Risk breakdown: - Collision risk: Calculated from other vessel distance and course - Weather risk: Calculated from wind speed and wave height - Fatigue risk: Detect captain concentration decline

AI Learning: - Past 3 years of operation data (normal navigation) - Past 1 year of near-miss reports (42 cases) - Learned veteran captain judgment patterns

After 4 months: AI model completed. In testing, could detect 39 of past 42 near-miss cases (93%) in advance

Phase 3: Decide - Presentation to Captain (2 months)

We presented AI judgment results in a form easy for captains to use.

New Bridge Screen:

Conventional (Captain monitors everything): - Radar screen - GPS screen - Weather information screen - Engine monitoring screen → Monitor 4 screens simultaneously (information overload)

New Screen (AI integration): - Central Screen: Comprehensive Risk Index - "Current risk: 28 (low)" - Risk breakdown: Collision 15, Weather 10, Fatigue 3 - Warning Screen (displays when risk 50+): - "Fishing boat ahead at 2km approaching, recommend: change course 5 degrees right" - "Wind speed rising to 10m/s predicted, recommend: reduce speed 20%"

Captain's Role: - Confirm AI suggestions - Final judgment by captain - "Change course as AI suggests" or "No, watch a bit longer"

Phase 4: Act - Autonomous Navigation Support (Ongoing)

Captain decides actions based on AI suggestions.

Autonomous Navigation Mode (Optional): - Can only be used when risk index is low (0-30) - AI automatically adjusts course and speed - Captain only monitors (can intervene anytime) - Reduces night burden

Chapter 4: The Power of Collaboration—Night Navigation After 6 Months

Results After 6 Months:

Night Operation Safety Improvement: - Near-miss incidents: 42 cases/year → 8 cases/year (81% reduction) - Of which at night: 32 cases → 5 cases (84% reduction) - Average collision avoidance reaction time: 3.5 min → 1.2 min (66% reduction)

Captain Burden Reduction:

Captain A's Testimony (Night watch 6 months later):

2:00 AM: Captain A on watch in bridge

AI Screen: "Current risk: 22 (low)" "No vessels nearby" "Weather: stable" "Fatigue level: 15 (normal)"

Captain A: "With AI monitoring, I feel secure. Previously I was tired watching 4 screens continuously, but now I just watch 1 screen."

2:30 AM: AI warning

Screen: "Risk: 58 (medium)" "Cargo ship ahead at 3km approaching" "Recommend: change course 10 degrees left"

Captain A: "Confirming AI suggestion... indeed, getting too close at this rate. Course change, understood."

Executes course change

3:00 AM: AI detects fatigue

Screen: "Fatigue level: 42 (somewhat high)" "Recommend: call first mate or switch to autonomous navigation mode"

Captain A: "Certainly, a bit sleepy... let's switch to autonomous navigation mode."

Autonomous Navigation Mode ON

AI automatically adjusts course and speed. Captain A takes light nap while seated in chair.

3:30 AM: AI wakes captain

Screen: "Risk: 65 (medium)" "Weather conditions deteriorating, wind speed rising to 8m/s" "Autonomous navigation mode cancel, requesting captain judgment"

Captain A: "Understood. Switching to manual."

Result: - Captain A's fatigue greatly reduced - Zero near-misses

Captain A's reflection: "Previously, the 4 hours from late night were truly painful. But with AI monitoring, I can nap with confidence. And I only need to concentrate when AI says 'human judgment needed.' 24-hour operations have become this easy..."

Business Results After 12 Months:

Safety: - Near-misses: 42 cases/year → 6 cases/year (86% reduction) - Marine accidents: Zero (continued) - Insurance premium: 4.8 million yen/year → 3.2 million yen/year (33% reduction, accident risk decreased)

Operational Efficiency: - 24-hour operation continuity rate: 92% → 98% - Captain resignation rate: 8%/year → 2%/year (work environment improved) - New captain training period: 2 years → 1 year (AI training support)

Financial Impact: - Investment: 48.4 million yen - Insurance premium reduction: 1.6 million yen/year - Crew hiring cost reduction: 4.8 million yen/year (resignation rate decreased) - Revenue increase from operational efficiency: 84 million yen/year - Investment recovery period: 5.5 months

Organizational Change:

Captain Meeting Evaluation:

Captain B (45 years old, 18 years experience): "At first I thought 'I'm worried about leaving it to AI.' But AI doesn't replace people, it assists. With AI monitoring the radar, I can concentrate on judgment."

Captain C (38 years old, 10 years experience): "When I was new, night watch was truly scary. I didn't know what to watch or how to judge. Today's new hires have AI teaching 'look here' and 'judge like this,' so training is faster."

Umino's Reflection:

"Before OODA × AI implementation, we felt '24-hour operations exceed human limits.' But we couldn't increase people either.

By implementing AI, we achieved role division where AI handles 'observation' and humans handle 'judgment.' Captains no longer need to monitor all information, only concentrate on 'information AI judged important.'

I understood OODA's true value is not 'speed' but 'adaptability.' Even when environment changes, with AI-human collaboration, we can respond immediately. 24-hour operations have become a sustainable system."

Chapter 5: The Detective's Diagnosis—Adaptation Means Thinking Calmly Amid Chaos

That night, I contemplated the essence of OODA and AI collaboration.

Asia Marine faced human limits with 24-hour operations. Fatigue, information overload, judgment delays.

However, by supporting the OODA loop with AI, humans could concentrate only on judgment. AI performs observation and presents situational assessment. Captains only need to confirm AI suggestions and make final judgments.

"Responsiveness means thinking calmly amid chaos. AI eliminates noise, people see only essentials. OODA creates optimal role division between people and AI."

The next case will also depict the moment when OODA creates adaptability.

"Don't increase people. Partner with AI. Observation by AI, judgment by people. OODA makes 24-hour operations sustainable"—From the Detective's Notes

Related Files