From an Hour of Uncertainty to a 5-Minute Decision
Agentic Decision Intelligence for feeder vessel operations — redesigning how operational knowledge becomes operational decisions.
An Hour Lost Before a Single Decision
When feeder vessel schedules are disrupted, teams spend significant time gathering, validating, and connecting information before deciding. That effort compounds into real cost:
- Decision delays — the vessel waits while the picture is assembled
- Operational inefficiency — effort duplicated across teams
- Increased cost — idle time and missed windows add up
- Customer uncertainty — commitments at risk downstream
Reduce the effort required to understand disruptions — so the decision cycle drops from ~42 minutes to 5, with humans still fully accountable.
More time deciding well.
Plans Rarely Survive Contact With Reality
Weather changes, ports congest, berths vanish, schedules shift. When they do, teams must answer three questions fast — yet most effort goes to understanding the situation, not deciding.
This was never a screen redesign — I set out to study how decisions get made, why they take time, and where information breaks down.
A Small Ship That Carries Big Consequences
Feeder vessels are smaller container ships that connect regional ports to major hubs, moving cargo to larger ports where it transfers onto bigger vessels.
Mother vessels are the international flights; feeders are the regional connectors.
No Single Stakeholder Sees the Whole Picture
Internal
- Vessel status
- Cargo priority
- Schedule commitments
- Resource constraints
External
- Weather
- Port congestion
- Berth availability
- Regulatory requirements
A berth is confirmed and everything looks on schedule — then weather turns, congestion climbs, and the allocation changes. The plan is no longer viable. What should happen next?
A Relay of Five, Before One Decision
It's fragmented information.
The Problem, Precisely Stated
High decision latency
High coordination effort
Increased operational risk
From "Show More" to "Decide Faster"
Better routing, tracking visibility, and notifications — on the assumption that more information would fix outcomes.
Information was already available. The real gap was turning information into decisions — so I reframed the goal around decision effort, not data volume.
| Business goal | UX strategy goal |
|---|---|
| Reduce delays | Reduce decision friction |
| Reduce cost | Reduce cognitive load |
| Improve reliability | Improve decision confidence |
| Improve efficiency | Improve information clarity |
I Stopped Asking, and Started Observing
"It depends." · "Every situation is different." · "We evaluate based on experience."
I learned what people said they did — not what they actually did.
I witnessed escalations, manual coordination, and WhatsApp workarounds in real environments.
What 42 Minutes Actually Looks Like
I mapped the current-state journey end to end — eight manual steps across disconnected systems, each with its own pain point, totalling 30–45 minutes before a decision could even begin.
Every Disruption Followed the Same Four Steps
Gather
Collect updates from weather, port, and tracking systems, plus calls and messages. Information exists — access is fragmented.
Connect
Mentally combine weather + port + cargo + vessel into context. Humans become the integration layer.
Teams Compare Consequences, Not Options
Acceptable
Manageable
Critical
Wait
Slow down
Divert
The Real Cause: Manual Context Creation
30 min understanding
10 min deciding
1 min understanding
4 min deciding
The Missing Decision-Assembly Layer
Detect → gather → connect → interpret → compare, automatically. Today, humans do all of this by hand.
Why AI — After Ruling Everything Else Out
Before landing on AI, I evaluated the realistic alternatives against one test: can it handle dynamic, ever-changing conditions — not just a fixed flow of static events?
Static automation
Rule engine
Works for predictable events — but disruptions combine unique, shifting variables. Rules break the moment reality changes. Too rigid.
API merging
Data integration
Unifies the data into one view — but a human still has to interpret it and decide. It moves data, not judgment. Doesn't reduce effort.
Do nothing
Status quo
The 42-minute cost and operational risk simply persist, unaddressed. Not an option.
Only AI can assemble context dynamically and evaluate shifting scenarios in real time — reasoning through conditions that change event to event, which is exactly what static approaches cannot do.
AI Assists. Humans Decide.
Having chosen AI, I anchored the whole design in one boundary: AI does the heavy lifting of assembling and evaluating — but the human keeps judgment, accountability, and the final call.
Reduce gathering
Compare faster
Surface impact
Preserve knowledge
Humans
- Accountability
- Judgment
- Risk acceptance
AI
- Context assembly
- Scenario evaluation
- Recommendation
Agentic Decision Intelligence
A decision-assembly layer that detects disruption, builds context, evaluates scenarios, and recommends — with a human firmly in the loop to approve and execute.
The Flow I Handed to Engineering
To take the concept into build, I authored a decision flow chart for the AI engineers and developers — defining what happens at each stage, and exactly where the human stays in control.
Decision points
Each stage has clear branch logic — e.g. low confidence routes back for more data before recommending.
Data contracts
Defined which systems feed each step, so engineering knew every input and output.
Human gate
Nothing executes without step 5 — approval is a hard stop, not a suggestion.
The Solution, Brought to Life
I translated the flow into key screens — from the operational cockpit to explainability, side-by-side scenario comparison, and the human approval gate.
Pressure-Tested With the People Who Live It
Captains, controllers, and fleet managers stress-tested the concept: Would you trust this? Where should AI stop? What must stay human?
AI is needed to
- Analyze
- Compare
- Recommend
AI is not expected to
- Approve
- Override captains
- Make safety calls
Confidence scores
Explainability
Traceability
Human approval
Same Disruption. A Different Outcome.
An investigation across a chain of stakeholders and disconnected systems before anyone could act.
Context arrives pre-assembled. The conversation starts at the decision.
The Challenge Was Never Routing
Operational
Faster response, less idle time.
Financial
Lower cost of delay.
Strategic
Higher network reliability.
Knowledge
Expertise scaled across the fleet.
By redesigning how operational knowledge becomes operational decisions, I reduced friction, improved response times, and preserved human expertise where it mattered most.