The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

Overview National International All initiatives

Whole of Australian Government Occupation Coding Service

https://www.abs.gov.au/about/key-priorities/whole-australian-government-occupation-coding-service

OECD analyst

23 Oct 2026

OECD analyst

23 Oct 2026

The Whole of Australian Government Occupation Coding Service uses AI/ML models and cloud compute to deliver fast, automated occupation coding to Australian standard classifications. It benefits agencies by providing consistent, high quality labour data. Built to replace manual coding which is slow and inconsistent, it solves fragmentation across government and enables real time, scalable coding for surveys, forms, and large datasets.

Initiative overview

Australia’s labour market intelligence depends on consistent occupation coding to the Occupation Standard Classification for Australia (OSCA) 2024, and the Australian and New Zealand Standard Classification of Occupations (ANZSCO) Australian Update 2022. Today, many agencies still rely on labour intensive manual processes and varied local practices, fragmenting data and slowing policy, research, and service delivery. The Whole of Australian Government (WoAG) Occupation Coding Service resolves this by providing a secure, AI/ML powered cloud service that automatically codes free text job titles and tasks in real time via API or at scale in batch.

Objectives. The initiative aims to (1) lift coding consistency and coherence across the public sector; (2) reduce effort and cost by maximising automated coding rates; (3) offer controllable and explainable models (hierarchical SVM as the primary method, with support for LLMs as appropriate); (4) provide secure, standards aligned access for agencies; and (5) improve the quality and timeliness of labour market data for decisionmakers and communities. The service is implemented on ABS-operated cloud infrastructure with segregated training/inference accounts, static models that do not learn from user inputs, and documented assurance under the Australian Government AI and security frameworks.

Current outcomes. The platform is live and already used by numerous agencies, enabling fast, consistent coding for surveys, forms, and enterprise datasets. Within the ABS, data from the 2021 Census of Population and Housing is being recoded to OSCA 2024 and the forthcoming 2026 Census will require substantially fewer clerical coders, freeing resources for targeted quality assurance and yielding significant productivity gains. The cloud-hosted service is able to serve multiple models efficiently, supporting single record real time calls and batch processing of millions of records.

The service has delivered major efficiency gains, with agencies coding 100,000 records in minutes rather than weeks and Census reducing temporary coding workforce needs while maintaining quality. Automated AI/ML models increase consistency across datasets and support adoption of updated classifications. Early use shows faster processing, improved comparability, and staff effort redirected from manual coding to quality assurance. Future impacts include broader uptake, reduced duplication, and more coherent labour market insights across government.

Other relevant details

Developing an AI/ML coding service required managing challenges common to emerging technologies. Ensuring highquality training data was critical, as model performance depended on accurate, updated inputs; this became more complex when classifications changed. Vendor changes and delays highlighted the need for strong continuity arrangements and early technical validation. Tools used to clean and prepare training data proved harder to stabilise than expected, showing that AI/ML components require early testing with real users to confirm usability and reliability. Key lessons included the importance of clear governance for model ownership, close collaboration between domain and technical experts, and transparent communication about model limitations. Success in initiatives like this depends on secure infrastructure, consistent assurance processes, and shared responsibility for quality throughout the AI lifecycle. Future evolution. The ABS will explore maturing the capability into a single enterprise service, including decommissioning legacy coders, and extending models to new topics. As adoption grows, the service is expected to become a core piece of national data infrastructure. Its future direction will focus on increasing adoption and integration of the service to support improvements to the collection and management of information within other government agencies. The capability will expand to support additional topics that also require consistent national classification, while continuing to improve the quality, transparency, and reliability of the AI models. As it becomes embedded across government, the service will gradually replace older, less consistent coding methods, reduce duplication and improve comparability across datasets. Its cloudbased design and structured assurance approach ensure the capability can evolve responsibly as technology and policy needs change, providing a sustainable foundation for evidence informed decision making.

About the policy initiative

Australia

Category:

AI policy initiatives, programmes and projects

Initiative type:

AI use cases/projects in the public sector

Status:

Active

Start Year:

2023

Target Sectors:

OECD AI Principles:

Other relevant urls:

—

AI in Digital Health

The Digital Health Agency’s AI initiative aims to improve how health information is used to support safer, more connected care for Australians. It helps clinicians and consumers by providing faster insights and reducing manual work. The initiative was created to address growing demand for timely, reliable health information and to ensure emerging AI tools are used safely and responsibly in the health system.

Start year: 2026