Australian Radio Network partners with Arinco on AI-created traffic bulletin workflow
The radio network wanted to bring bulletins in-house without employing more journalists.
Australian Radio Network (ARN) has partnered with Arinco to develop a multi-agent AI platform to create traffic bulletins for its radio network.
The network’s more than 3,000 traffic reports were supplied by an external provider, but ARN wanted greater ownership and the ability to create new commercial opportunities. To move in-house, ARN estimated it would need to employ 24 full-time employees at a cost of approximately $3.8 million a year.
A proof of concept was trialled across six markets in four weeks, before developing the capability into a governed platform for live broadcast operations over a further 14 weeks.
“Arinco understood what we were trying to achieve and brought the technical capability and delivery approach to make it happen quickly,” said Mat Williams, head of technology at Australian Radio Network.
In the pilot, the system completed approximately 500 generation runs, with around 190 bulletins approved by ARN operators and approximately 170 delivered to station playout systems.
Arinco focused on improving how the platform handled changing traffic information, tuning bulletin timing for different voices and giving stations control over local pronunciations.
The team also had to ensure operators could readily review and edit bulletins before broadcast and their feedback was central to progressing from proof of concept to live operations.
Agents perform distinct roles within the bulletin creation process. Some analyse traffic incidents and road flow, then pass relevant information to an agent that drafts the bulletin. Another reviews the script against the source information and assesses its suitability for radio.
Information passes between these stages, allowing the script to be revised before audio is generated. The station’s team then receives the recording, script and supporting source information together, enabling operators to review the content, make changes and approve it for broadcast.
Each station can configure its own voice, delivery style, pronunciations, priority roads and branded opening and closing lines. Weather and fuel information can also be incorporated where relevant, while new stations can be added through configuration rather than requiring a separate build.
“The result gives us much more control over how the traffic product sounds, operates and develops commercially, while keeping our people responsible for what goes to air,” said Williams.
Looking ahead, the radio network may extend the system into broader information script generation and content research support.
Marc Tricksey, national sales director at Arinco, said the project showed the value of using agents in two ways — within the ARN platform and embedded into how its team delivered the project.
“That agentic delivery approach helped us move quickly from experimentation into a governed, live capability, while our consultants and ARN’s teams kept the judgement, quality and controls around it,” he said.