Machine Learning & Artificial Intelligence

Case studies from government agencies and jurisdictions around Australia focussing on the use of machine learning and artificial intelligence in developing government services.

You will hear:

  • how machine learning was implemented to allow better predictions of taxpayers default
  • how AI, machine learning and image recognition is giving people with disabilities and impairments more freedom
  • how a new employee has reduced call wait times and improved staff productivity
  • how automating records classification has improved the confidence level

Case studies:

  • Not just the HOW and WHAT, but also the WHY: using machine learning to predict tax debt – Queensland Treasury
  • Making it easier to get around: autonomous transport – IBM and South Australian government
  • Queensland Government’s first digital employee – ‘Russell’ – Queensland Health
  • Better records classification – Human Rights Commission


Not just the HOW and WHAT, but also the WHY: using machine learning to predict tax debt

In 2017, the Queensland Office of State Revenue (OSR) began a 3-year transformation program. The aim was to provide next-generation tax and revenue management capabilities that are client-centric, digitally-enabled and data-driven–to improve outcomes for taxpayers, staff, the Government and the Queensland community.

Learn how OSR is leveraging machine learning as part of this program, including to predict taxpayers who may be at risk of becoming debtors, before they default.

Insights provided through the machine learning application, not only identify the how and the what around who will default, but also the influencing factors for why taxpayers fail to pay their tax on time.

This innovative solution has helped OSR better understand individual taxpayer’s circumstances and needs, identify taxpayers who may be at risk of becoming debtors before they default. It has also helped to implement proactive intervention strategies to prevent taxpayers from becoming debtors; and deliver more timely revenue collection – ensuring that the Government can meet its commitments to Queenslanders.

Making it easier to get around - autonomous transport

Major shifts in our social spaces, such as an ageing population and growing social awareness, have highlighted the need to make sure people of all ages and capacities have access to public services.

In January 2019 the South Australian Government launched the ‘First-mile-last-mile’ initiative.

This project consisted of a 6-month trial of Olli–a self-driving electric shuttle, and Matilda–an interactive, smart transit hub. This presentation will show how the use of IBM capabilities, in collaboration with SAGE Automation and Local Motors, enabled the provision of a wide range of travel services to those struggling with the accessibility of public transportation.

Features such as an audio and surveillance system, image recognition, sign language proficiencies, machine learning, and interactive competences, developed a personalised and unique passenger experience for all, including the older Australians and people with disability.

The success of this project shows the powerful role digital transformation can play in creating a much more autonomous and accessible world for everyone.

Queensland Government’s first digital employee–‘Russell’

eHealth Queensland is responsible for ensuring the smooth operation of digital solutions and technologies to Queensland Health employees across the state. The organisation has experienced a 19% growth in IT support requests, bringing the total amount of requests to just over 1.75 million per year.

Through implementing ground-breaking technological innovation and introducing Queensland Government’s first digital employee ‘Russell’, our automated virtual agent, eHealth Queensland can address pain points, and provide customers with a positive experience regardless of the communication and engagement channel. Since his implementation, Russell has contributed to the reduction in call volumes and queuing, and improved staff productivity across Queensland Health. He does this by enabling them to continue working until he phones them back through our call back functionality.

Russell is also used to reset forgotten or expired passwords–nearly 20% of call volumes, log job details to hold the customer’s place in a queue, can facilitate a customer’s update on their most recent job, or provide custom announcements based on the location the caller is phoning from. Russell might be the first product in our automation and innovation journey, but he is definitely not the last.

Better records classification

The Australian Human Rights Commission (the Commission) has started implementing an Electronic Document and Records Management System (EDRMS), in line with the Australian Government’s Digital Continuity 2020 Policy.

Mindful of mixed satisfaction with the traditional EDRMS, we identified the need for an intuitive system tailored to our work processes, which removed staff from records classification decisions.

The Commission partnered with RecordPoint to develop RADICAL (Records and Document Innovation & Capture – Artificial Learning).

RADICAL uses RecordPoint’s Records365 Classification Intelligence Engine to auto-classify records using machine learning. SharePoint Online serves as RADICAL’s user interface.

To develop RADICAL, we supplied RecordPoint with a large dataset of records classified against the AFDA Express and Commission authorities. We worked closely with their developers to build and refine a machine learning model. The resulting model classifies records with a confidence level of 80%.

To develop our SharePoint Online interface, we consulted extensively with our staff and other agencies using SharePoint Online.

A key lesson learned with RADICAL is that machine learning is not a panacea for information management problems. RADICAL classifies records based on their content, but does not incorporate context, which can lead to inaccuracies. We continue to work with RecordPoint to resolve these issues.

Date and time



Bradman Theatrette



Presentation type

Case studies