ERP2 & OBUs: What Singapore’s Rollout Teaches AI Ops

Singapore’s move to fully switch to ERP2 from Jan 1, 2027 is one of those policy changes that looks “transport-only” on the surface—until you notice what it really represents: a national-scale technology adoption programme with deadlines, incentives, enforcement, and public trust requirements.

That matters for anyone running operations in Singapore, especially in logistics and supply chain. If you’re trying to implement AI route optimisation, demand forecasting, or warehouse automation, you’re dealing with the same challenges ERP2 is tackling: hardware/software rollout, user acceptance, data privacy, exception handling, and change management.

The core story from the news is straightforward: Parliament is debating a Bill to make on-board units (OBUs) mandatory for Singapore-registered vehicles because ERP2 uses satellite-based (GNSS) location rather than gantries. But the practical lessons for businesses go further than “install the device.”

ERP2 is a reminder that tech adoption doesn’t succeed because the technology is impressive—it succeeds because the rollout is designed for real behaviour.

What’s changing with ERP2 (and why it’s a big deal)

Answer first: ERP2 replaces gantry-based charging with satellite-based charging, which requires vehicles to have an OBU so the system can determine the vehicle’s location for charging.

According to CNA’s report (Feb 3, 2026), Acting Transport Minister Jeffrey Siow said ERP2 will be more precise in targeting congestion and will remove the need for expensive physical gantries. It also enables a fairer approach—charging can be spread across multiple points instead of a single choke point.

A few concrete rollout details are worth paying attention to because they’re exactly how good enterprise AI programmes are run:

• Deadline-driven change: Full switch on Jan 1, 2027.
• High adoption before enforcement: About 93% of vehicle owners have installed the OBU.
• Final reminders + limited-time free installation: From Feb 15, invited owners who haven’t installed get a final reminder and 3 months to install for free.
• Post-free window installation fees: S$35 for motorcycles and S$70 for other vehicles.
• Fallback pricing to keep the system operable: From Jan 1, 2027, those without OBUs pay flat fees per ERP operational day: S$3 (motorcycles) and S$10 (all other vehicles).
• Governance and enforcement: Tampering/misuse can lead to penalties up to S$20,000, up to 2 years’ jail, or both.

In other words: it’s not “here’s new tech, good luck.” It’s a structured adoption plan.

Why mandatory adoption works (and where businesses get it wrong)

Answer first: Mandatory adoption works when the programme combines a clear “why,” a realistic transition path, and strong governance—otherwise people bypass the system.

Most companies get AI adoption wrong in a predictable way: they treat it like a tool purchase (“we bought an AI forecasting platform”) rather than a system change (“we changed how decisions get made”). ERP2 is being rolled out like a system change.

Here are the parallels I’d copy directly into a logistics or supply chain AI roadmap.

1) Make the “why” operational, not inspirational

ERP2’s “why” isn’t vague digital transformation language. It’s practical:

• ERP1 infrastructure is aging and costly to sustain.
• ERP2 can target persistent congestion hotspots more precisely.
• It removes gantries and enables finer-grained pricing.

For an AI logistics implementation, your “why” should be equally concrete:

• “Reduce late deliveries from 8% to 4% by improving route plans.”
• “Cut picking time per order by 12% using warehouse slotting models.”
• “Lower safety stock by 10 days without increasing stockouts.”

If you can’t state it in a number, teams will treat it as optional.

2) Build a transition that respects current workflows

A detail in the article that’s easy to miss: motorists can choose not to have a display screen, to keep the experience closer to ERP1.

That’s excellent change design. It acknowledges user friction and offers a path that preserves what people already know.

In supply chain AI, the equivalent is:

• Keep planners in their familiar interface (Excel/ERP screens) while AI outputs feed in as recommendations.
• Start with decision support before going fully automated.
• Provide a manual override path with logging (so exceptions become training data).

3) Use incentives first, penalties last

Singapore used free installation and reminders, then introduces fees and flat-rate charging for non-compliance.

For business AI adoption:

• Incentive: faster approvals, fewer admin steps, better SLA outcomes.
• Penalty (soft): new process becomes the default; old process requires extra justification.
• Penalty (hard): deprecate old systems and remove access.

If you try to enforce without support, people create shadow workflows.

The data question: privacy, governance, and trust

Answer first: Any location- or behaviour-tracking system must have strict governance, because adoption collapses when people believe data will be misused.

CNA reported that MPs raised concerns around privacy of vehicle movement data (e.g., WP MP Dennis Tan). That concern is not a side issue—it’s central to whether people trust satellite-based systems.

For companies deploying AI in logistics and supply chain, privacy and governance show up in different forms:

• Driver tracking and route optimisation data
• In-warehouse worker productivity telemetry
• Supplier performance scoring
• Customer order pattern modelling

Here’s what “good governance” looks like in a business context (and it’s the same logic ERP2 needs):

• Purpose limitation: Collect data for a specific operational purpose (ETA accuracy, route adherence), not “because we can.”
• Retention rules: Decide how long raw logs are kept versus aggregated metrics.
• Access controls: Operations teams don’t need raw individual traces by default.
• Auditability: Every model-driven decision (reroute, priority change) should be traceable.

One stance I’ll take: if your AI programme can’t clearly answer “who can see what, and why,” you’re not ready to scale.

ERP2 as a real-world example of AI in logistics

Answer first: ERP2 is essentially a national, usage-based optimisation system—exactly the kind of feedback loop AI operations teams try to build inside companies.

Even though ERP2 isn’t branded as “AI,” it behaves like the foundation layer many AI systems depend on:

• Continuous sensing: GNSS location as a stream of operational signals.
• Pricing/charging rules: A policy engine that reacts to time and place.
• Hotspot targeting: The concept of applying interventions where congestion persists.

In the AI dalam Logistik dan Rantaian Bekalan series, we often talk about four pillars:

1. AI mengoptimumkan laluan pengangkutan (route optimisation)
2. Automasi gudang (warehouse automation)
3. Ramalan permintaan (demand forecasting)
4. Keberkesanan rantaian bekalan (end-to-end efficiency)

ERP2 connects strongly to pillar #1. If Singapore can redesign pricing signals to influence traffic flow, your company can redesign operational signals to influence delivery performance.

Practical example: using “charging logic” thinking for fleet optimisation

ERP2’s idea of distributing charges across multiple points is similar to distributing operational constraints across a route plan.

Instead of one big rule (“avoid Orchard Road”), you design several smaller constraints:

• Avoid certain road segments during known peak windows
• Penalise routes that historically create late arrivals
• Add a cost for left turns or U-turns where safety incidents cluster

That’s how real optimisation improves—by turning messy reality into weighted costs.

A 6-step rollout checklist you can steal for AI adoption

Answer first: Treat your AI tool rollout like ERP2: phased onboarding, clear exceptions, user choice, and enforcement-ready governance.

Use this as a playbook for AI projects like route optimisation, warehouse picking AI, or demand forecasting.

1. Define the policy and success metrics
   • Example: “Reduce cost per stop by 6% within 90 days.”
2. Instrument the system (data capture) before automation
   • If data quality is weak, models will be ignored.
3. Start with a default experience that feels familiar
   • AI suggestions inside existing planner workflows.
4. Create a free/low-friction onboarding window
   • Training, templates, assisted setup, office hours.
5. Design the fallback mode
   • ERP2 uses flat fees for those without OBUs; you need a workable manual process for exceptions.
6. Lock governance early
   • Access, retention, audit logs, and what counts as misuse.

If you do steps 1–4 well, you rarely need to “fight” adoption.

What to watch between now and Jan 1, 2027

Answer first: The most important period is the transition—how exceptions, privacy concerns, and enforcement are handled will determine public confidence.

A few signals from the article suggest what’s coming next:

• The Bill includes offences for tampering/modifying OBUs and for unauthorised services on the OBU.
• The government said it won’t introduce distance-based charging in the immediate term, focusing first on a smooth experience similar to ERP1.
• There are exemptions (e.g., vintage/classic vehicles), and different rules for foreign vehicles (flat fee option), while Malaysian taxis must install OBUs for enforcement.

That “no immediate distance-based charging” decision is a classic stabilisation move: don’t change the policy model while users are still adapting to the platform change.

Businesses should copy this. When you implement AI forecasting, don’t also overhaul procurement policies and supplier terms in the same quarter. Change is additive; so is resistance.

Next steps for logistics teams adopting AI in Singapore

Singapore’s ERP2 rollout is a case study in how to scale technology with public scrutiny: clear rules, staged onboarding, and a serious approach to governance.

If you’re running logistics, fleet, or supply chain operations, the actionable takeaway is simple: AI adoption succeeds when it’s managed like infrastructure, not like software. Treat data capture, trust, and fallback processes as first-class requirements.

If you want a practical plan, I can help map an “ERP2-style rollout” to your business—covering AI route optimisation, demand forecasting, and warehouse automation in a way that teams actually adopt. What part of your operation has the biggest gap right now: routing, warehouse throughput, or forecasting accuracy?

Source article: CNA report published Feb 3, 2026. (No external links included per publishing rules.)