The Australian Broadcasting Corporation has described an AI workflow that gives regional radio reporting an additional digital output. Local bulletins can be transformed into website material, allowing journalism made for a scheduled audio programme to reach audiences online and in more localised forms.
ABC says the production prompt runs three or four pages and was co-authored by engineering and newsroom staff. It builds on the broadcaster’s AIAssist work, automatic transcription and a grounded model developed with Google over its archive and production context.
The editorial boundary is explicit: a person reviews and publishes the result. ABC positions the system as removing repetitive conversion work rather than producing autonomous news. The organisation is also introducing Claude for staff behind mandatory training described as a licence to use the tool.
No accuracy rate, correction volume or time saving has been published. The workflow will need continuing tests across accents, local names, legal risk and changes between spoken and written style. Archive grounding also requires clear access control and rules for rights-sensitive material.
