A "platform-plus-agile" approach has enabled the Victorian Department of Human Services (DHS) to deliver more than 30 business systems at a speed never thought possible in a government setting, said Steve Hodgkinson, CIO at the Victorian Department of Health & Human Services.
At Microsoft's Creating a Digital Difference Summit, Hodgkinson said the agile methodology cannot be applied alone in government because "you hit a brick wall as soon as something is proven to be good", whether it's around procurement, security, or integration with legacy environments.
The platform-plus-agile approach is about starting with the production platform, then building a "minimum visible product" on top of it, he explained, which is followed by iterative improvements.
"You don't have to build all the basic technology. You don't have to go out and do unnecessary procurement. We can actually just use things the way they were designed to be used, get on with it, and do things," Hodgkinson said.
This approach has allowed the Victorian DHS to deliver the L17 family violence referral and triage system within six months. The system now allows police to capture more accurate information about individual cases and ensure a faster response to family violence.
Previously, when Victorian police officers suspected domestic violence during house visits, they would fax over referrals to social services agencies. Around 70,000 faxes were being sent annually, amounting to 150,000 referrals in some cases, Hodgkinson said.
Multiple attempts had been made to modernise the 10-year-old system, he added, but "the wheels always fell off" due to difficulties around getting complex processes approved across more than 50 different government agencies, including Victoria Police and the Department of Justice and Regulation.
But the interest in developing a better system "revved up" following recommendation by the Royal Commission into Family Violence, Hodgkinson said. It was the murder of 11-year-old Luke Batty by his father in 2014 that exposed the need for better coordination in sharing information within Victoria Police.
That same year, Victoria Police acknowledged in a report [PDF] that it was struggling to meet the demands of the public, and that inadequate investment in technology had "left Victoria Police in the 20th century". The state government has since boosted its investment in technology for police and protective services personnel.
From June to December 2016, L17 went from concept to "minimum visible product" using in-house resources, as well as known and trusted platforms such as Microsoft Azure and Oracle Siebel Open UI.
"The notion of, with minimum documentation possible, documenting use cases into storyboards and that kind of thing, in order to build something really quickly, and then get feedback through sprint sessions from users, was very, very powerful," Hodgkinson told ZDNet.
"It meant that it was possible to take a co-design, problem-solving approach to it, rather than having to spend a whole lot of time iterating documents to get them perfect and signed off. That was important because Victoria Police has very stringent security requirements and so it was important to build trust with them and iterate solutions with them so they felt that there would be adequate security around their data."
It was a combination of using platforms the DHS already had at its disposal -- with robust procurement arrangements already in place -- as well as in-house development teams that made it possible for the department to proceed at a pace never before achieved.
"You can only do that if you're empowered by the knowledge of what your platforms can do, which means you've got to have people that know the platforms, know what they're good for, know what works, know what doesn't work, know what's hard, knows what's easy. When you put a person like that together with a business user, they very quickly work out what would be a minimum visible product," Hodgkinson told ZDNet.
It's all about finding the "sweet spot" between what would be useful and what can be developed, he added.
"You're trying to empower people to have and exercise common sense. And usually common sense gets lost in government because of the time that elapses between someone having a good idea and that good idea turning into something that can be executed. It can take years," Hodgkinson said.
Hodgkinson additionally said that the government needs to see itself as a consumer of modern application environments, rather than thinking everything needs to be specially made for government.
"We don't have to go to market for every single thing; what we have to do is get value for money for our taxpayers," he said.
"Rather than saying 'how can I specify an application and go out to market and buy it', a better conversation these days is 'how can we satisfy this use case in the platforms that we already have'.
"If we had used the traditional method, we still would be arguing about the business case and the specifications, and if we had gone to market, we would probably still have been in procurement negotiating the contract."
L17 has gone through numerous iterations -- big and small -- since its launch in December. The DHS has created additional APIs to ensure interoperability between L17 and Victoria's core child protection case management system, and is currently finalising the process that would enable Victoria Police to access the system to close that feedback loop.
More than 30 business systems have been delivered in the last 12 to 18 months using the platform-plus-agile approach, Hodgkinson said, including the Victorian Housing Register, which allows state residents to apply for social housing on their smartphones.
"That's another transformational use case entirely built on the Microsoft stack, which leverages the Microsoft .Net development environment, our own devops capabilities, and Microsoft Azure, and was implemented in a similar iterative co-design way and is available on a smartphone through the fact that applications built in modern environments now render to smartphones with minimal additional work," Hodgkinson told ZDNet.
"[Previously] the idea at the time was that it could not be done on the smartphone because it's too complicated. But we were able to build it without much extra effort and cost, and it turns out that a large number -- if not the majority -- of the users have completed their application with a smartphone because that's the device they all have.
"If applications are built in these modern environments -- whether it's Salesforce or the Azure stack -- then they just inherit functionalities and features and benefits of that environment. If we had done it in a traditionally way, we probably wouldn't have even designed it to work on a smartphone. It wouldn't have been in the requirements."
Hodgkinson said the Victorian DHS will continue to refine its approach to building new systems, but the next phase is all about using more advanced analytics and reporting processes. The department is currently looking into using predictive analytics in areas such as child protection risk assessment.
"There's some very exciting work being done in the department on the use of predictive analytics to give or inform risk assessments, helping child protection workers understand complex case files. If you're a new child protection worker and you're presented with a complex family where there might be 10 to 15 to 20 years of historical data about that family from a large number of people, a large number of case information, how do you make sense of that?" he said.
"By applying artificial intelligence onto that, there is a way of synthesising and pulling out of complex case files core information that a caseworker needs to know."
The IT industry can confidentially email the government about its grievances.
After evaluating a range of GST collection models, the Productivity Commission has concluded that the legislated model used for high-value goods should also be used for low-value goods.
Agile project management: The smart person's guide (TechRepublic)
This is the go-to guide on Agile project management for project managers, business leaders, developers, project and product teams, consultants, stakeholders, and students.
IBM's machine learning algorithm achieved 33.8 percent accuracy on image recognition, beating Facebook and Microsoft. It could speed up deep learning training from "days or hours to minutes or seconds".
The DECODE function offers a handy way to make data easier to analyze and more understandable when you share it with others. Here's a quick review of how it works.