AirFish Singapore–Batam: Lessons for AI Logistics Teams

ST Engineering says its AirFish “wing-in-ground” craft can travel at three times the speed of existing marine craft—and a Singapore–Batam trip could drop to about 25 minutes versus the ~45 minutes travellers are used to. That sounds like a transport story, but it’s also a very Singapore story: ambitious engineering, tight regulation, and a business case that lives or dies on operations.

For anyone following our “AI dalam Logistik dan Rantaian Bekalan” series, this announcement matters because it exposes the part most teams underestimate: when you add a faster asset, the bottleneck simply moves. The winning move isn’t just buying speed. It’s designing the workflow around it—planning, risk assessment, scheduling, pricing, staffing, maintenance, customer comms. That’s where AI in logistics and supply chain stops being a buzzword and becomes a daily advantage.

The CNA report (Feb 2026) lays out the basics: ST Engineering AirX signed an agreement with regional ferry operator BatamFast to lease an AirFish and operate it on the Singapore–Batam route, targeting 2H 2026, pending regulatory approvals. Ticket prices aren’t final, but BatamFast’s general manager expects them to be higher than current ferries because the AirFish carries up to 10 people (including crew), which increases unit cost.

Below is the bigger lesson: AirFish is a case study in what it takes to operationalise advanced tech—and how AI tools can help Singapore businesses do the same across logistics and supply chain.

Why AirFish matters for logistics (beyond passenger travel)

AirFish isn’t “just a faster ferry.” It’s a different operating model.

A conventional ferry system is optimised for:

• Larger capacity per trip
• Predictable docking and turnaround
• Mature maintenance supply chains
• Familiar regulatory categories

AirFish changes several of those assumptions at once:

• Speed increases, but route constraints (traffic lanes, speed limits) still apply
• Capacity decreases, raising the stakes on load factor and pricing
• It doesn’t require conventional take-off/landing infrastructure, which can open new route designs
• It runs under a classification/certification process (ST Engineering expects classification by mid-2026, working with Bureau Veritas since 2024)

Here’s the operational reality I’ve seen across “new asset” deployments: the tech works, but the organisation struggles to decide how to run it. That’s where AI becomes practical.

In supply chain terms, AirFish is like introducing a premium, high-speed lane into a network. If you don’t re-plan inventory, staffing, pricing, and customer expectations, you’ll pay more for speed without capturing value.

The real bottleneck: turnaround time, not cruising speed

A 20-minute improvement (45 → 25 minutes) is meaningful—especially for business travellers and premium tourism. But the customer experience is shaped just as much by:

• Check-in and security steps
• Boarding queues
• Baggage handling rules
• On-time departure discipline
• Weather and sea-state decisions
• Maintenance readiness and spare parts availability

What AI can do here (and what it can’t)

AI won’t change physics or regulations, but it can shrink variability:

1. Demand forecasting (ramalan permintaan): Predict which sailings will fill based on day-of-week, holidays, events, and historical travel patterns.
2. Dynamic scheduling: Adjust frequency and departure times to match predicted peaks—without burning crews or causing maintenance slippage.
3. Queue optimisation: Use real-time passenger flow data to staff counters and gates correctly.
4. Predictive maintenance (penyelenggaraan ramalan): Detect early signals from sensor data and logs to avoid last-minute cancellations.

A useful stance: speed is a product feature; reliability is the product. AI is usually a reliability engine.

Unit economics: small capacity forces precision

BatamFast was direct: tickets will “certainly” cost more than existing ferries due to the AirFish’s smaller capacity and higher unit costs. That’s not a problem—unless the operator guesses.

With up to 10 people including crew, revenue per trip is constrained. That forces sharper decisions on:

• Price tiers (standard vs premium)
• Bundling (fast-track check-in, lounge access, hotel tie-ups)
• Which passenger segments to prioritise
• How many trips per day are operationally safe and profitable

AI pricing and revenue management—applied realistically

When people hear “AI pricing,” they imagine surge pricing. I’m not a fan of blunt surge models in regulated, reputation-sensitive markets like Singapore.

A better approach is constraint-based revenue optimisation:

• Set guardrails (max/min price, fairness rules, transparent fees)
• Forecast demand by segment
• Allocate seats per fare bucket
• Learn from no-show rates and rebooking behaviour

Even simple models can beat spreadsheet guessing if you maintain good data hygiene: bookings, lead time, cancellations, weather impacts, and operational disruptions.

Partnerships and regulation: your AI rollout has the same shape

This AirFish deployment is a partnership story:

• ST Engineering AirX provides the craft
• BatamFast operates and sells the service
• Authorities provide route and regulatory approvals
• Classification partner (Bureau Veritas) supports certification readiness

That pattern mirrors enterprise AI in logistics:

• A vendor supplies the model/platform
• Ops teams run it in production
• Compliance/security teams set boundaries
• External stakeholders (customers, regulators, auditors) expect accountability

A practical “approval pathway” for AI in logistics teams

If you want AI adoption to move faster than your internal politics, design an approval pathway upfront:

1. Define the risk assessment (the same phrase BatamFast used)
   • What decisions will AI influence?
   • What’s the worst credible outcome?
   • What controls prevent it?
2. Choose a certification standard (internal)
   • Data retention rules
   • Access controls
   • Model monitoring and incident response
3. Pilot with measurable KPIs
   • On-time performance
   • Cost per trip/order
   • Customer complaints per 1,000 passengers/orders
   • Maintenance downtime

AirFish is waiting on approvals before launch. Most AI projects stall for the same reason: no one agrees on what “safe and ready” looks like.

Route expansion: why longer lanes benefit more (and how AI helps)

BatamFast noted that time savings are limited on short routes due to restrictions, and that benefits increase on longer routes. They’re also exploring other destinations such as Bintan and Tioman Island.

This is a network design point that logistics teams should internalise:

Faster assets create more value when the network lets them stay fast.

In supply chain and transport planning, that means:

• Fewer stop-start constraints
• Better slot availability
• Fewer handoffs
• Stable operating conditions

AI route optimisation (pengoptimuman laluan) for mixed fleets

If you introduce a new class of asset (like AirFish), your planning becomes a mixed-fleet optimisation problem:

• Which routes suit which vehicle class?
• What’s the minimum viable frequency?
• How do you cover disruptions?

AI-based optimisation can help you simulate scenarios quickly:

• “If we assign AirFish to Route A, what happens to load factor and spare capacity?”
• “If weather grounds AirFish 8% of days in monsoon season, what’s the best fallback plan?”
• “How many spare parts should we stock in Singapore vs Batam to hit 98% dispatch reliability?”

The point isn’t perfect prediction. It’s faster, defensible decisions.

People also ask: Is AirFish a plane or a boat?

It’s best described as a wing-in-ground (WIG) craft that uses the aerodynamic ground effect to glide a few metres above the water. Operationally, that puts it in a space between aviation and marine operations—one reason certification and route approvals matter.

For business teams, the category matters because it affects:

• Safety procedures
• Crew training requirements
• Maintenance regimes
• Insurance
• Customer communications (expectations and comfort)

What Singapore operators can copy from this playbook

AirFish is the headline, but the transferable playbook is the process.

1) Start with service design, not the technology

Speed is easy to market. The service is harder:

• What is the promise—“25 minutes gate to gate” or “25 minutes water time”?
• What happens when there’s a delay?
• How do you rebook customers instantly?

AI can support this with automated disruption workflows and customer messaging, but the promise must be defined first.

2) Treat operational data as an asset from day one

New services often drown in manual exception handling. If you want AI to help, capture clean operational data early:

• Trip logs (departure/arrival, delays, causes)
• Maintenance events and parts usage
• Weather/sea-state decisions
• Customer demand signals (searches, abandoned carts, lead times)

No data, no forecasting. No forecasting, no profitable frequency.

3) Build a “human-in-the-loop” operating model

For high-stakes transport, fully automated decision-making is a bad idea.

A solid AI operating model looks like:

• AI recommends
• A duty officer approves
• The system records rationale
• Monitoring detects drift and anomalies

That’s how you keep speed and trust.

Memorable rule: Automation without accountability becomes chaos at scale.

What to do next if you’re leading AI for logistics in Singapore

AirFish is scheduled for operations in the second half of 2026, subject to approvals. Between now and then, the winners will be the teams that treat advanced mobility as an operations discipline—supported by AI.

If you’re building AI in logistics and supply chain (ramalan permintaan, pengoptimuman laluan pengangkutan, automasi gudang, and end-to-end visibility), copy the structure:

1. Pick one operational bottleneck (forecasting, scheduling, maintenance, or customer comms)
2. Define “ready for production” with risk controls
3. Pilot with real KPIs and real constraints

The next wave of competitive advantage in Singapore won’t come from having the newest tech. It’ll come from running it reliably, day after day—while everyone else is still debating.

What would change in your operations if your transport time dropped by 40%, but your capacity per trip also dropped by 70–80%? That’s the kind of trade-off AI is best at managing.

Source referenced: https://www.channelnewsasia.com/singapore/st-engineering-airfish-water-skimming-wing-in-ground-vessel-batam-ferry-route-5903436