Executive summary

Use-case map cells by maturity tier
Singapore container terminals · 8 process areas x 4 lenses = 32 cells · Nov 2026
Proven live, at scale, owner-reported 3 remote cranes, yard cranes, AGV routing Scaling live in part of the operation 12 berth windows, stacking, trucks Experimental PoC, MoU or trial only 10 AI incident response, lashing robots No Singapore case found tools exist, no result here 7 dwell forecasts, planner co-pilots
Source: [Kira estimates] from the use-case map in chapter 2

The answer in three lines

Planning is already algorithmic. PSA states that AI runs across berth planning, container stacking, predictive maintenance and the routing of its driverless vehicles [MG Jul 2025]; over 300 AGVs worked 11 berths in 2025 [AL Mar 2025].

Resilience is the open frontier. In October 2026 PSA and Cisco started a proof of concept for AI that watches the terminal's applications and networks and recommends fixes before outages spread [Cisco Oct 2026].

The operator keeps most of the gain; carriers keep the fuel saved by arriving just in time, and hauliers the empty trips avoided [SWP OptEVoyage; PSA OptETruck].

Value at stakeWe size AI at about S$90-215 mn a year by 2030, roughly 4-8% of terminal operating cost [Kira estimates].
A Japanese or Korean technology firm should enter through resilience, maintenance and document tools, tested in the Tuas Phase 2 Living Lab, not by competing with PSA's own planning systems.
AL Mar 2025 · Cisco Oct 2026 · MG Jul 2025 · PSA OptETruck · SWP OptEVoyage · Kira estimates (full titles in the source index)

From vessel arrival to gate-out, the planning, fleet and IT tasks are both data-heavy and digitisable; lashing and mooring are neither

Exposure differs by task, not by company: the same terminal is fully automated in the yard and still manual at the ship's side.

Terminal tasks by how far AI can take them
Singapore · KIRA scores 1-5 · bubble = relative cost weight · Nov 2026
Automate now Data first Mechanise later Tools exist, small prize How far the work can be digitised (1-5) Information intensity (1-5) Berth and vessel planning Yard planning and stacking AGV fleet orchestration IT operations, incidents Gate and documents Equipment maintenance Quay crane operation Mooring, marine services Lashing and unlashing
Source: [Kira estimates]

Why planning leads and the ship's side lags

Planning is already software. Berth windows, stack positions and vehicle routes are computed decisions, and PSA says AI is embedded in each of them [MG Jul 2025].

The ship's side is not. Lashing specialists still fix and remove twist locks by hand next to the quay cranes; PSA describes robot lashing as exploratory and human-in-the-loop [MG Jul 2025].

IT is now a task in its own right. When the terminal operating system stops, the cranes and vehicles stop; that is why the newest AI project is about outages, not productivity [Cisco Oct 2026].

Scoring ruleDigitisability = share of the task a system can do or check without a person; information intensity = how much the task depends on data.
Start where the work is already data and the decision is made by software: planning upgrades, fleet control and IT operations.
Cisco Oct 2026 = Cisco Newsroom · MG Jul 2025 = Maritime Gateway · Kira estimates = KIRA analyst triangulation