In 2024, for the second consecutive year, our Poland A2 motorway operator, AESA (Autostrada Eksploatacja SA) was recognised for its digital innovation.
Up against 45 other companies (including blue chips like Orange, BNP Paribas and Allianz), AESA took the prize in the 10th CIONET Digital Excellence Awards* for its Digitized Call Centre. This system not only identifies vehicle locations but also provides real-time traffic and toll information. In this article, Arkadiusz (Arek) Sitek, IT Director and Change Manager, tells the story of how their award-winning SOMA (Highway Operation and Monitoring System (System Obsługi i Monitorowania Autostrady) system came about.
We employ five hundred and eighty people to operate and maintain the 255 km long A2 motorway from Świecko on the German border to Konin in the East. We have five operational maintenance centres and one control room for the whole motorway, with five patrols out every day 24/7. Each year we deal with 22,000 events ranging from road traffic accidents to spillages, repair works and more. The planning, coordination, communication and analysis involved in all this is substantial.
Back in 2017, we decided that with a digital approach, we could radically improve the efficiency and reliability of our operations. Our aim was to coordinate fieldwork, report incidents, inform drivers/users about incidents, plan roadworks and manage repairs to damaged infrastructure – all with zero paperwork – we needed an Event Database.
In the road maintenance industry, 100% off-the-shelf solutions rarely fit the operational and reporting needs of our kind of business. Our first attempt at implementing this project used a completely off-the-shelf solution. After a year of trying, we went ‘back to the drawing board’ and decided to start again, combining off-the-shelf software with our own in-house programs.
The first version of the system was developed and used between 2017 and 2019. Leveraging the experience gained during this period, we decided to rebuild the entire software. The second version has been in use since 2020, and we have adopted a constant improvement and development approach, adding new functionalities every few months. While the system was fully functional and usable before 2023, by that year it had become significantly more refined, complete, and impressive, giving us the confidence to enter the Digital Excellence Awards competition organized by CIONET.
From prototype to go-live in 16 months
Our IT and Operations Centre teams worked closely together to develop the concept and the operational algorithms. We prioritised creating and testing a prototype, enhancing it as we went along, and holding weekly meetings with the programming team. Sixteen months later, the system went live. It integrates with navigation, weather, signage, and radio systems, an incident image bank, customer helpline, and social media.
Today, all reports and necessary source documents, including patrol logs, are generated digitally from start to finish. This has significantly reduced the reporting time from several days to just one. It enables real-time notification of road conditions and detours to drivers, and data security mechanisms have made it a reliable source of information for regulatory authorities. From the commercial point of view, we can now predict and manage patrol response times and roadwork coordination to ensure we comply with Service Level Agreements (SLA).
One unexpected benefit of the innovation is that, thanks to reliable information about road conditions now being available, the use of navigation applications by customers has increased to 87% (compared to 69% before the project), with 78% of users using them regularly (compared to 50% before the project). This enhances road safety for both our customers and employees. Another achievement, and one that was quite surprising given the context, was the ability to reduce annual patrol driving from 3 million to 2.7 million kilometres. We managed this 300,000 km reduction despite simultaneously seeing a 25% increase in interventions. This operational efficiency improvement translates directly into lower fuel costs and a positive reduction in our CO2 footprint.
At a glance: impacts of the Event Database:
- Information about a road event is transmitted to external systems (WAZE, Yanosik, variable message signs) in less than 2 minutes.
- Information about a possible SLA travel time breach is displayed to the control room operator in less than 10 seconds.
- All highway events are electronically recorded, including detailed information about signage changes during event management.
- Other reports related to current highway maintenance are generated in less than 1 minute.
- Field employees are notified of their proximity to an event 1,000 meters ahead in less than 2 seconds.
- Information about the direction, speed, and location of the vehicle is updated (milepost, object) every 5 seconds.
Digitized Call Centre
Once the Event Database was up and running, we quickly realised how it could become a building block for more digital innovation. This time the idea was to automate our customer helpline using AI and Large Language Models (LLM). A voice/chatbot system would integrate with internal and external systems, enabling real-time and accurate information to be shared, including traffic, fees, and disruptions. The goal was to increase efficiency (handling a larger number of inquiries at lower costs) and to provide customers with the most current, contextual information.
We started with a needs analysis, gathering data from the hotline, call transcriptions, and the Highway Management Centre, identifying the most frequent questions and problems reported by customers. This gave us the data to develop a range of service scenarios. From there, we began preparing training data, drafting 600 sentences, and expanding them with an additional 3,000 sentences generated using AI. When it came to testing the model, we used employees from various departments and local LLM models to simulate different types of users. As a result of the testing process, we introduced some new enquiry paths and expanded the data set for model training to 4,500 sentences and 68 intent labels. The final design step was to introduce a mechanism for assigning cities and phrases according to the direction of travel and road junctions.
Seven months later, the Digital Call Centre went live. Replacing human operators has resulted in an 85% reduction in costs associated with the external hotline and a 50% increase in call-handling capacity. We believe the project is the first of its kind in the industry. Its modular design allows for easy expansion both in terms of functionality and support for subsequent areas. It works in Polish and English and has the potential for many more languages. The team have called it SOMA, Highway Operation and Monitoring System (System Obsługi i Monitorowania Autostrady) and you can see it in action, here!
