willyam - stock.adobe.com

Services Australia taps AI in document processing

Services Australia and Capgemini have developed an AI engine to classify and extract information from claim documents in a pilot project

This article can also be found in the Premium Editorial Download: CW Asia-Pacific: CW APAC: Buyer’s guide to data management

Services Australia and Capgemini have developed a new artificial intelligence (AI) engine in a pilot project to classify and extract information from documents submitted by citizens who are applying for Centrelink, Medicare and Child Support payments and services. 

Built on open-source natural language processing and optical character recognition (OCR) tools, the AI engine has enabled Services Australia to cut down the time it takes to process citizen documents – which can go up to 25,000 per day – from weeks to seconds.

Olaf Pietschner, CEO and managing director of Capgemini in Australia and New Zealand, said through the use of progressive scale expansion networks and regression-based neural networks, the project has achieved an accuracy rate of 95%.

Work on the project started in August 2019 when Capgemini was asked by Services Australia to demonstrate how intelligent technology could be used to classify documents uploaded by citizens in their payment claims.

At the time, the backlog for classification was thought to be around three million documents, driving the need for an accurate way to classify the documents quickly, Pietschner told Computer Weekly.

“We’ve been helping Services Australia work through that backlog, streamline the submission process and reduce the wait time for citizens,” he said, adding that the containerised technology could also be adapted to the needs of other government agencies and private sectors such as financial services.

Lysandra Schmutter, vice-president for Capgemini’s federal government business, said the project has come into its own during the Covid-19 pandemic, which has increased demand for citizen services.

“It has condensed the time it takes for a citizen to launch documents with Services Australia to a case being given to a case worker to assess,” she added.

Read more about AI in Australia

Pietschner stressed that the goal of the project was not to reduce headcount. Rather, he said it was to cut down the amount of manual work, enabling Services Australia staff to focus on higher value activities such as customer service to ensure support and payments to citizens are timely.

“That’s really the major outcome – to streamline the interaction with citizens and allow people to focus on things where you need value judgment, not checking documents,” he said.

Schmutter said amid the Covid-19 pandemic, and as demonstrated in the recent Budget, the Australian government has been accelerating digital transformation efforts to provide more services to citizens online.

“We observe that they’re very focused on modernising legacy infrastructure, and collaborations like what we have seen with Services Australia is absolutely what governments are looking for,” she added.

Under the 2020-2021 Budget, the government has allocated A$539.6m to complete the fourth stage of the seven-year build of Services Australia’s Welfare Payment Infrastructure Transformation programme.

The fourth phase will see the development of modern payments and decision-making platforms, including an entitlement calculation engine that will speed up claims processing for Australians and ensure accuracy of payments.

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management