DHS aims for AI-driven service delivery future

With virtual assistants already proving useful in aiding workers at the Department of Human Services, chief information officer Gary Sterrenberg says attention is now turning to the technology's potential to provide “formidable service delivery”.

What began with an internal virtual assistant dubbed ‘Roxy’ to help staff deal with complex questions about the processing of student, carer and age pension claims and some 100,000 legislative rules has spread beyond compliance to areas of customer service.

In the space of a year, three similarly personified bots have arrived with this focus in mind, which Sterrenberg says is part of the department's multi-pronged approach towards virtual assistance.

“We focus virtual assistance in three main category areas: One is customer and channel management, the second is on service delivery, which is really about claims and service processing, and the third is around back office functions, where you can automate using deep learning techniques and algorithms related to the type of work we do in the backend,” he told iTnews.

One of the new assistants is ‘Sam’, a public-facing chatbot that was launched in July to answer general questions customers might have about the students, trainees and families payments pages of DHS’ website.

The department has also introduced ‘Oliver’, another public-facing assistant for student claims, but which is only available through the myGov online services portal.

Sterrenberg said Sam and Oliver were the department's first virtual assistants to cater to customer and channel management.

“The intention is to ensure that we use conversational UI to improve the correctness of data that citizens or recipients of claims provide when they are answering information online," he said.

“That allows us to get cleaner claims, which makes the processing more likely to be straight-through.”

Unlike Sam, Oliver offers a personalised experience to the claimant by tailoring its questions based on what it already knows about the individual.

“What Oliver does is understand – because it’s in a protected, authenticated environment – the circumstances of the person claiming and is able to adjust the questions to reflect only the information we don’t know about the student,” Sterrenberg said.

Oliver has only been live a couple of months, but Sterrenberg says the results have already been “extremely positive”, with around a 92 percent satisfaction rating given through a Facebook-style thumbs up, thumbs down rating system.

He said there have close to 650,000 interactions across DHS' bot army so far, and he expects this to climb to close to a million by the end of the year.

“The early signs are that this could be a significant enhancement to our already significant capabilities in servicing Australians that have one vulnerability or another,” Sterrenberg said.

Cortana, Watson and TensorFlow

Sitting behind Roxy, Sam and Oliver is Microsoft’s Cortana technology, a partnership that has seen DHS invited as one of only four organisations worldwide – and the only public sector administration – to partner on a project with Microsoft referred to as Toronto.

“We are working in a development sense with Microsoft to see how we can move machine learning to the next level, which we all believe is going to be where there’s self-learning," Sterrenberg said.

“It’s not just the ability to learn intents and having those intents based on probabilistic algorithms, it’s really about the machine itself to use broader datasets to create meaning and learning through self-learning.”

But while Cortana is utilised for virtual assistants, it is just one of three cognitive platforms that are being used across the deaprtment.

Sterrenberg said IBM's Watson platform was still “being used in some actuarial models around the disability sector”, despite the department casting doubt on its future application earlier this year.

“We see it as a potential player, particularly for unstructured, longitudinal datasets,” he said.

However, it remains to be seen whether Watson will play a role in the federal government’s much-hyped Nadia virtual assistant, with the department hinting at recent budget estimates that IBM been replaced by another provider.

The department is also using Google’s open-source TensorFlow software for robotic process automation.

Looking ahead

DHS is in the early stages of tinkering with different incarnations of virtual assistants in its technology innovation centre, including a small physical robot known as George.

“George is starting to explore solving some significant challenges that we have in our face to face servicing,” Sterrenberg said.

“One of these is violence. George is able to detect violence in a crowd, and that give us signals to be able to help our staff avoid those situations.”

George is able to communicate in 32 different languages, which Sterrenberg says will help the department serve the 25 percent of its customers that are new to the country.

And while George currently takes the form of a robot, Sterrenberg says the technology needs further testing before being introduced into shopfronts, and won't necessarily be the final form that it takes.

“The use of form factor is just one of interest really to see what’s possible, to see computing in a very different way,” he said.

“George allows us to change the conversation both with the customers we engage with and with our own staff to say there could be a different way, and hence that’s how were approaching things in DHS.”