How is AI being applied to Electronic Toll Collection?

AI has been part of tolling for years, notably in applications such as image review. What is changing is the pace of innovation and the range of opportunities generative AI has opened up. From identifying vehicles to helping employees resolve customer disputes, AI is becoming an increasingly important part of the road charging operating model.

Its clearest contribution is operational efficiency. Better image recognition can improve licence plate read rates and reduce manual review. Enforcement automation supports fraud detection and case prioritisation, while workflow automation can reduce processing errors and improve transaction accuracy. These applications can lower operating expenses (OPEX), while revenue assurance also helps protect toll income.

Looking ahead, we expect AI’s benefits to extend into revenue optimisation and potentially lower capital expenditures (CAPEX). Predictive analytics could support pricing decisions and help anticipate congestion and incident risks. Better use of existing assets could reduce the need for unnecessary sensors and overbuilding, or support lighter infrastructure over time. AI agents could also improve service for toll customers.

Where AI is gaining ground?

To understand this evolution, PTOLEMUS analysed 36 initiatives and solutions trialled or implemented by DOTs, public agencies and private concessionaires in North America and Europe. This review covers tolling solution vendors that represent more than 50% of the vendors, providing a snapshot of current activity rather than a measure of industry-wide adoption.

Roadside systems account for the most observed initiatives, with 10, followed by customer service centres with 8. Operational back offices and enforcement each account for 7, while commercial back offices account for 4. These counts highlight activity across the entire tolling value chain, not just at the roadside.

The examples span several operational needs. Several players, such as Emovis and Tecsidel, use AI-assisted optical character recognition (OCR) on challenging number plate images; Yunex Traffic’s YuTraffic Fusion uses road network data to optimise traffic flow in real time; and Neology’s NeoZone uses AI to combine multi-sensor data and process transactions more effectively.

In customer service, Pennsylvania Turnpike’s Miles provides round-the-clock virtual assistance with tolls, E-ZPass accounts and travel conditions. In enforcement, Kapsch TrafficCom’s HoTCap uses AI-powered vehicle fingerprinting to identify vehicles by their overall appearance, strengthening identification beyond traditional number plate recognition.

Live applications today, broader capabilities tomorrow

Within the initiatives assessed, existing applications already include image processing, vehicle tracking, revenue leakage detection, charging and billing checks, chatbots, fraud detection and trip reconstruction. AI is therefore addressing both the accuracy of toll transactions and the work required to manage them.

Pilot applications extend into traffic and revenue forecasting, satellite-based journey processing, dynamic toll pricing, roadside infrastructure monitoring and agent training. The emphasis is shifting from recognising and detecting to predicting, recommending, and optimising. However, pilot activity should not be confused with widespread operational deployment.

We can actually identify 3 waves of implementation, where:

• Wave 1 began when toll road operators adopted computer vision and machine learning for free-flow tolling, using ANPR, automated image review, axle counting, and fraud pattern detection to improve vehicle identification accuracy and transaction processing.
• Wave 2 builds on the data generated in Wave 1, introducing advanced analytics such as dynamic pricing and customer behaviour prediction.
• Wave 3 is now emerging to optimise toll infrastructure, using AI to manage assets more efficiently and helping suppliers determine the best actions and timing for system updates and maintenance. It includes predictive maintenance for toll equipment based on sensor data. It aims to avoid costly failures and defer CAPEX.

The table below details PTOLEMUS’ analysis of AI implementation by building block and deployment status.

GenAI’s next role: A copilot for employees

One emerging opportunity is to embed GenAI within customer service and enforcement workflows. Repetitive, policy-driven contacts such as missed tolls, invoice queries, account changes, payment questions and disputes offer a practical starting point. While it is just starting, companies such as Telepass and ViaPlus have documented its use.

GenAI could summarise interactions, retrieve relevant policies, prioritise cases, suggest next steps and draft responses. Translation and routing support could further reduce administrative work. The intended result is shorter handling times, less after-call processing and more time for agents to focus on complex cases.

Should « human-in-the-loop » remain the rule?

Human agents should retain final approval for sensitive decisions. The objective is to make operations leaner and employees better informed—not to replace meaningful judgement with automatic acceptance of an AI recommendation.

However, AI agents could improve customer service in a number of cases where humans have obvious limits.

For example, by providing 24/7 service. Roads continue to be used even when customer service agents prefer to go to sleep. As the negative health impacts of nighttime work become evident, AI agents will complement staff efficiently.

Similarly, what appeared impossible yesterday is now possible. Once agents are used, multilingual customer service appears reachable. For example, in countries with a large share of international road traffic, such as Belgium or the Netherlands, several language options could be provided in addition to the national languages. Viapass and RDS could include German, Polish and other languages frequently spoken by Heavy Goods Vehicle (HGV) drivers. Similarly, US agencies that struggle to find a business case for customer service in Spanish can have agents automatically translate user-facing workflows.

AI’s next chapter in tolling will therefore be about more than reducing manual work. The opportunity is to combine stronger revenue protection, better use of infrastructure and enhanced customer service. All with measurable benefits and clear human accountability.

Sources:

PTOLEMUS analysis of selected AI initiatives and solutions reported by technology providers and toll operators..

Organisations identified: Emovis, Kapsch TrafficCom, PA Turnpike, Tecsidel, ViaPlus, Yunex.

To learn more about how new technologies are shaping the Tolling Solutions market, download our report here.  Contact Alberto Lodieu if you would like to learn more about implementing AI in your business or find out how the competitive landscape is evolving.

PTOLEMUS intellectual property. Feel free to use it by quoting PTOLEMUS Consulting Group as the source.