ERP2 & Mandatory OBUs: Lessons for AI Operations

Singapore has put a date on it: ERP2 becomes the full system from 1 Jan 2027, and Parliament is debating a Bill that would make on-board units (OBUs) mandatory for all Singapore-registered vehicles. As of now, about 93% of vehicle owners have installed the OBU, and LTA is already moving to “final reminders” and paid installations after a grace period.

That’s transport policy on the surface. Underneath, it’s a clean case study of how Singapore rolls out nationwide technology: tighten the rules, simplify the workflows, protect edge cases, and keep the user experience close to what people already accept.

If you’re following this “AI dalam Logistik dan Rantaian Bekalan” series, you’ll recognise the pattern. AI adoption inside a company looks a lot like ERP2 adoption on the roads: the tech matters, but governance, incentives, data handling, and change management decide whether it works.

What’s changing with ERP2 (and why it matters)

Answer first: ERP2 shifts road pricing from gantry-based detection (ERP1) to satellite-based positioning (GNSS), which is why an in-vehicle OBU becomes the enforcement and billing anchor.

ERP1 is reaching end-of-life after almost three decades. The government’s message is blunt: keeping obsolete infrastructure alive has become “unsustainably challenging and expensive.” That’s not just a transport problem; it’s the same reason companies replace legacy systems that were “fine” until maintenance became the business risk.

ERP2 in one practical sentence

ERP2 lets Singapore price congestion with finer granularity—more precise charging points, smaller charge amounts across multiple locations, and fewer physical gantries to install and maintain.

For operations people, this is the key move: from fixed checkpoints to continuous context. In supply chain terms, it’s like moving from periodic stocktakes to real-time inventory visibility.

A detail businesses should pay attention to: the “no big bang” promise

Acting Transport Minister Jeffrey Siow stated the intent is a smooth transition so ERP2 feels as similar as possible to ERP1 at the start. Crucially:

• Distance-based charging won’t be introduced immediately.
• It will be studied later, after people get used to the new system.

This is a textbook rollout strategy. You don’t start with the most disruptive pricing model. You start by stabilising the platform and adoption.

In AI terms: don’t begin by automating the hardest end-to-end process. Start with high-confidence use cases (forecasting assistance, routing suggestions, exception triage), then expand once trust and data quality improve.

Mandatory OBUs = a blueprint for enterprise AI adoption

Answer first: The move to make OBUs mandatory shows that standardisation beats optional participation when a system depends on consistent data capture.

ERP2 needs a common “device layer” in vehicles so the system can charge correctly and enforce fairly. Businesses face the same reality with AI tools: if each team uses different data definitions, disconnected apps, and ad-hoc prompts, you get inconsistent outputs and endless exceptions.

Here are the parallels worth stealing for your own AI transformation.

1) Adoption isn’t a marketing problem; it’s an operations problem

LTA’s approach mixes reminders, free installation windows, then paid installation:

• After a final reminder, owners have three months to install for free.
• After that, installation costs S$35 (motorcycles) and S$70 (other vehicles).
• From 1 Jan 2027, vehicles without OBUs pay flat fees per ERP operational day: S$3 (motorcycles) and S$10 (all other vehicles).

That is behavior design.

For businesses adopting AI in logistics and supply chain, the equivalent isn’t “encourage staff to use the chatbot.” It’s:

• Define the default workflow (where AI is embedded)
• Make the non-standard path slightly painful (extra approvals, manual reporting)
• Make the standard path frictionless (templates, integrations, single sign-on)

2) “Similar user experience” is a serious strategy

OBUs sparked feedback, and the system was adjusted—such as letting owners opt out of a display screen for a more familiar experience.

I’m opinionated here: most digital transformations fail because leaders treat UX as decoration. For frontline operations—dispatchers, warehouse supervisors, fleet managers—UX is throughput.

If you’re rolling out AI tools (route optimisation, demand forecasting, warehouse slotting), design for:

• Minimal extra clicks
• Outputs that fit existing decisions (not new dashboards nobody checks)
• Clear “why” explanations for recommendations

What ERP2 teaches about AI in logistics and supply chain

Answer first: ERP2 shows how to operationalise data-driven decisions at national scale—exactly what AI aims to do inside logistics networks.

Singapore is effectively upgrading from “infrastructure-defined constraints” (gantries) to “data-defined constraints” (location signals + policy rules). In logistics, AI does the same: it turns planning and execution into a loop where data continuously updates decisions.

Route optimisation: from static rules to dynamic pricing signals

ERP2’s promise is more precise targeting of congestion. When congestion pricing becomes more granular, logistics routing also changes:

• Delivery windows become cost variables, not just time variables
• Dispatch decisions need near-real-time context
• Fleet utilisation is affected by micro-charges across multiple points

In practice, companies can mirror ERP2 logic internally by using AI to:

• Re-route based on live traffic + service-level commitments
• Cluster deliveries to reduce stop-start inefficiencies
• Shift non-urgent jobs outside peak pricing periods

Demand forecasting and capacity planning: governance matters more than models

ERP2’s rollout is backed by legislation: rules for tampering, misuse, and unauthorised services.

This matters because AI forecasting can be “tampered with” too—just less visibly. If teams can change inputs, override assumptions, or selectively report results, the model becomes political.

A practical governance checklist for AI forecasting in supply chain:

1. Define input ownership (who owns sales history, promo calendars, returns)
2. Log overrides (who changed the forecast, when, and why)
3. Measure error honestly (MAPE/WMAPE by category and horizon)
4. Audit data pipelines (missing values, late feeds, versioning)

The two biggest concerns: privacy and obsolescence (and the business equivalent)

Answer first: MPs raised two concerns—data privacy and technology obsolescence—and these are the same two issues that derail enterprise AI programmes.

Privacy: movement data vs operational data

MP Dennis Tan raised concerns about privacy of vehicle movement data collected by the OBU.

For businesses, the analogue is sensitive operational telemetry:

• Driver location and behaviour
• Warehouse worker productivity traces
• Customer order patterns

If you want adoption, treat privacy as a design constraint, not a legal footnote. What works:

• Collect the minimum required for the business outcome
• Keep retention periods explicit (and short where possible)
• Separate personal identifiers from operational events
• Provide clear internal policies on who can access what

A quotable rule I use: “If you can’t explain the data policy in two minutes to a frontline supervisor, it’s not ready.”

Obsolescence: devices age; architectures should not

MP Tin Pei Ling asked what measures will ensure OBU technology doesn’t become obsolete.

In AI, your model might still be fine while your tooling becomes the bottleneck—integrations, workflow engines, data warehouses, MLOps pipelines. Future-proofing is mostly about architecture:

• Prefer modular components (swap models without rewriting the workflow)
• Standardise APIs and event formats
• Avoid hard-coding business rules inside the model
• Build monitoring from day one (drift, latency, cost)

A practical playbook: adopting AI tools the “ERP2 way”

Answer first: The ERP2 rollout suggests a five-step playbook businesses can copy to deploy AI tools in logistics and supply chain with fewer surprises.

1) Set a date, then work backwards

ERP2 has a clear cutover date: 1 Jan 2027. Your AI initiative also needs deadlines that force integration work to finish.

2) Make participation default

Mandatory OBUs create standardisation. For AI, embed tools into:

• TMS (Transport Management System) dispatch screens
• WMS task assignment
• Procurement reorder approvals

3) Keep the first release familiar

No immediate distance-based charging mirrors a sensible AI approach:

• Start with decision support before full automation
• Keep human overrides—then measure them

4) Design for edge cases, not just the average

ERP2 includes exemptions (e.g., vintage and classic vehicles) and special rules (e.g., Malaysian taxis required to install OBUs).

In logistics, edge cases are where your margin disappears:

• Cold chain shipments
• Customs exceptions
• High-value deliveries
• Returns spikes after seasonal campaigns

5) Put teeth in the rules

The proposed law makes unauthorised tampering an offence, with serious penalties for deliberate misuse (up to S$20,000, up to two years’ jail, or both).

Your company doesn’t need legal penalties, but it does need guardrails:

• Role-based access to models and prompts
• Approval workflows for changing critical assumptions
• Audit trails for automated decisions

What to do next (especially if you run fleets, deliveries, or ops)

Singapore’s ERP2 shift is a reminder that technology rollouts succeed when they’re treated as systems, not gadgets. A satellite-based road pricing platform needs devices, policy, enforcement, support, and a staged transition. AI in logistics and rantaian bekalan is exactly the same.

If you’re planning AI adoption this quarter, don’t start by buying tools. Start by mapping the workflow you want to standardise—routing, forecasting, warehouse execution—and decide where AI should sit inside it. Then lock the data and governance down early.

The question worth sitting with: if your organisation had to make AI usage “mandatory” in one workflow next year, which workflow would you choose—and what would you need to fix first so people don’t hate it?

Source: https://www.channelnewsasia.com/singapore/erp-obu-mandatory-jan-1-parliament-jeffrey-siow-5903661