ERP2 Mandate: What It Teaches About AI Ops in SG

Singapore’s full switch to ERP2 on Jan 1, 2027 isn’t just a transport upgrade. It’s a very Singapore move: replace aging infrastructure, standardise the tech, then use better data to manage a scarce resource—in this case, road space.

If you work in logistics, fleet operations, or supply chain, you should pay attention. Not because you’re suddenly in the tolling business, but because ERP2 is a public, high-stakes example of the same pattern businesses are living through with AI: more sensors, more data, more automation—and new rules about how that data can be used.

The ERP2 rollout also lands at a practical moment. Early 2026 is when many Singapore teams are finalising budgets and tooling decisions for the year. The question isn’t “Should we adopt AI?” It’s “Where do we mandate it internally so operations don’t fracture into five different systems?”

What’s changing with ERP2 (and why the mandate matters)

Answer first: ERP2 replaces gantry-based charging with satellite-based location charging and requires an on-board unit (OBU) for most Singapore-registered vehicles—because the system can’t work reliably at scale without standardised hardware.

Under the Bill debated in Parliament (Feb 2026), OBUs become mandatory for Singapore-registered vehicles ahead of the full switch on Jan 1, 2027. Key points from the reported details:

• ERP2 uses GNSS/satellite technology, not physical gantries, to determine chargeable road usage.
• About 93% of vehicle owners have installed the OBU so far.
• From Feb 15, owners who were invited but haven’t installed will receive a final reminder, with 3 months to install for free.
• After that, installation fees apply: S$35 (motorcycles) and S$70 (all other vehicles).
• From Jan 1, 2027, vehicles without OBUs pay flat fees per ERP operational day they use the roads: S$3 (motorcycles) and S$10 (others).
• Tampering, unauthorised modification, or unauthorised services/advertising on OBUs can carry penalties up to S$20,000 and/or up to 2 years’ jail for serious cases.

Why did Singapore push a mandate? Because optional adoption creates gaps. You end up maintaining two charging systems (old and new), plus handling exceptions manually. That’s expensive, messy, and unfair.

This is exactly what happens in businesses that “encourage” AI tools but never standardise anything: some teams automate, others don’t, and leaders get a dashboard that looks confident while the underlying data is inconsistent.

ERP2 is a live case study in data-driven operations

Answer first: ERP2 is really a real-time operations optimisation system—pricing road usage based on where and when congestion happens. That’s the same logic AI brings to routing, capacity planning, and warehouse labour scheduling.

Acting Transport Minister Jeffrey Siow described ERP2 as “more precise and finer” in targeting congestion, and “fairer” because charges can be spread across multiple points rather than concentrated at one gantry.

From a supply chain lens, that is familiar:

• Congestion is a capacity constraint.
• Pricing (or prioritisation rules) is how you allocate constrained capacity.
• The better your location/time data, the more targeted your intervention can be.

In logistics terms: ERP1 vs ERP2 is like barcode scans vs GPS + telematics

ERP1’s gantries are like fixed checkpoint scans. Useful, but limited.

ERP2’s GNSS model is closer to what fleet telematics already does: continuous location awareness enabling finer-grained decisions.

What AI adds on top is prediction and optimisation:

• Predict congestion impacts on ETA
• Recommend departure times
• Suggest alternate routes
• Flag delivery sequences that reduce late stops

ERP2 doesn’t promise distance-based charging immediately (the minister said it won’t be introduced “in the immediate term”), but the platform is designed to support more nuanced models later.

That’s another lesson for businesses rolling out AI: start with an experience people can tolerate, then progressively add complexity after trust and habits form.

The part businesses miss: mandates are about interoperability, not control

Answer first: Mandates are unpopular, but they’re often the only way to make a system interoperable—so insights are comparable, enforcement is possible, and exceptions don’t swallow operations.

Parliamentarians raised issues that any operations leader should recognise:

• Obsolescence risk: What if the OBU tech becomes outdated?
• Privacy and surveillance concerns: What happens to movement data?
• Equity: Who bears the cost burden—especially lower-income families, private-hire drivers, and delivery riders?

Translate that to AI in supply chain:

• What if our AI vendor’s model or API changes and breaks our workflows?
• Are we collecting worker productivity data in a way that crosses a line?
• Are we pushing efficiency targets that increase cost or stress for frontline teams?

The point isn’t to avoid mandates. It’s to mandate responsibly.

A practical “OBU policy” equivalent for AI tools in operations

If I were writing the internal version of the ERP2 Bill for a company rolling out AI in logistics and rantaian bekalan, I’d include:

1. Standard toolchain for core processes

   • One source of truth for orders, inventory, and route plans
   • Approved AI tools for forecasting, routing, and exception triage

2. Data governance rules that are specific

   • What data is collected (GPS pings, scan events, driver notes)
   • How long it’s kept
   • Who can access it
   • When it can be used for performance management

3. Tamper and misuse controls

   • Audit logs
   • Role-based access
   • Clear consequences for bypassing controls (similar in spirit to OBU tampering offences)

4. An upgrade path

   • Vendor portability plan
   • Model monitoring and re-training schedule
   • Budget line for integration work (because “AI subscription” is never the whole cost)

Mandates are easier to accept when people see the guardrails.

How ERP2 maps to AI use cases in logistics & supply chain

Answer first: ERP2 shows the value of location intelligence + policy rules. In business, AI turns that same raw data into operational decisions: better routes, better demand forecasts, and fewer manual exceptions.

Here are four direct parallels that fit this post’s series theme (AI dalam Logistik dan Rantaian Bekalan):

1) AI route optimisation under real congestion constraints

ERP2’s whole purpose is to manage congestion hotspots. Logistics teams can treat congestion as a variable cost and constraint.

AI route optimisation tools typically combine:

• Historical travel times (by time-of-day)
• Real-time traffic signals
• Stop constraints (delivery windows, service time)
• Vehicle constraints (capacity, cold chain)

What works in practice: start by optimising just one slice—say, morning routes in the CBD—then expand. This mirrors Singapore’s plan to keep the “experience similar” during transition.

2) AI demand forecasting to reduce last-minute trips

Congestion charges (and congestion itself) punish reactive operations. Better forecasting reduces:

• Expedited deliveries
• Partial-load trips
• Emergency replenishments

Even a modest forecast improvement can cut “extra runs,” which tends to show up quickly in fuel, labour, and vehicle utilisation.

3) Warehouse labour scheduling and slotting

ERP2 is about better targeting. Warehouses have the same problem: labour and dock doors are limited; peaks are predictable; exceptions disrupt everything.

AI-driven scheduling can:

• Predict inbound/outbound volume by hour
• Recommend shift allocations
• Identify which SKUs should move closer to packing lines before peaks

The key is the same as ERP2: precision beats blanket rules.

4) Exception management: from alerts to actions

ERP2 OBUs can support more than charging (CNA previously reported features like flood alerts on OBUs). In business ops, AI should also move beyond dashboards.

A good exception system doesn’t just say “late.” It says:

• “Late because of congestion on PIE; recommend reroute via X.”
• “Customer at stop 7 is a repeat no-show; propose reschedule window.”
• “Driver has 3 stops with tight windows; swap stop order.”

This is where AI tools for operations actually earn their keep.

What to do now (Feb 2026): a short checklist for ops leaders

Answer first: Use the ERP2 transition as a prompt to standardise your own AI adoption—before you end up with fragmented data, duplicated work, and compliance risk.

If you run a fleet, delivery operation, 3PL, retail logistics team, or supply chain function in Singapore, here’s a practical checklist you can finish this quarter:

1. Map your “gantries” (fixed checkpoints) vs your “GNSS” (continuous signals)

   • Where are you still relying on manual status updates or periodic scans?
   • Where could GPS/telematics + automation remove blind spots?

2. Pick 1-2 AI use cases tied to a hard KPI

   • Route adherence / on-time-in-full (OTIF)
   • Cost per drop
   • Forecast accuracy
   • Warehouse pick rate and error rate

3. Define the policy before the tooling

   • Who owns the data?
   • What’s private vs operationally necessary?
   • What decisions can AI recommend vs auto-execute?

4. Budget for integration and change management

   • Most failures aren’t model failures—they’re workflow failures.
   • Train supervisors, dispatchers, and planners on “how we work now.”

5. Set an internal mandate date

   • Not a vague “rollout.” A date when a process switches.
   • Provide a grace period and support, like the free OBU installation window.

A line I come back to: if it’s optional forever, it won’t become operational.

ERP2’s bigger signal: Singapore will keep standardising tech

Singapore is switching to ERP2 because the old system is reaching end-of-life and maintaining it became “unsustainably challenging and expensive.” That’s the boring truth behind most tech transformations—public or private.

For businesses, AI adoption is heading the same way: less experimentation for its own sake, more standard platforms with governance. The teams that win won’t be the ones with the most pilots. They’ll be the ones who turn pilots into reliable processes.

If you’re building capabilities in AI dalam logistik dan rantaian bekalan, treat ERP2 as a reminder that technology rollouts succeed when three things are aligned: hardware (or systems), rules, and trust. Which part of your operation is still missing one of the three?