DTA CEO suggests AI usage to reduce 'white-collar effort' in APS

The head of the Digital Transformation Agency has pointed to artificial intelligence’s potential to “lift productivity” across the federal government’s program delivery amid a “really strong fiscal constraint”. 

The agency’s CEO Chris Fechner indicated that AI had already “proved itself to be more than equal” in carrying out certain functions such as “taking minutes, doing actions, [and] drafting first versions of documents”

Speaking on a Public Sector Network webinar, Fechner said: “There are many things that take an enormous amount of white-collar effort.

“Artificial intelligence has proved itself to be more than equal in reducing complexity.

“One of the great things about artificial intelligence, I believe, is that it should reduce the cost and time of doing things that aren’t particularly fun or interesting, allowing us to spend more time on things that have a greater impact.”

Fechner, whose agency led the federal government's six-month trial of Microsoft Copilot last year, noted mounting pressure on the Australian Public Service (APS) to contain program costs, especially given the compound impact of project failures.

“There is a really strong fiscal constraint, and making sure that our programs around delivery and policy are promoted means that our internal costs need to be heavily scrutinised,” he said.

“I think the success of all the initiatives we bring forward needs to have a stronger guarantee.

"We can’t have failures, because a failure has double the impact: you don’t get what you need, and you’ve spent the money.”

As such, using AI “to provide the right augmentation”, alongside skills and capabilities, could play a key role in lifting productivity across the public sector.

Fewer $100 million projects

Fechner, who was named DTA CEO in 2021, also pointed to federal procurement’s role in inhibiting productivity, noting that Commonwealth projects average “generally between two and three years” in length.

“We think it should be half that,” he said. “We think it should be even less than half that in some cases.

“It shouldn’t be $100 million projects, it should be $20 million projects. The more we can actually realise value in shorter periods with lower costs, the more likely we are to be successful.

“We also have evidence that the longer a project runs, the less likely it is to realise its originally intended benefits.”

“It’s also about simplifying things,” he added. “Instead of looking at 100 things, we look at 10 things and make decisions on those so that we can move to value quicker.

GovERP “not fallen into a hole”

His comments come off the back of the government’s efforts to find reusability within the failed GovERP project, after the program was found to have “produced little of functional value”.

An expert panel did note there was “strong demand” for access to its repository of process maps, designs and patterns.

Commenting on the failed program, Fechner said the project’s legacy had indeed not “just fallen into a hole”.

“Many of the core processes were captured out of there, and Services Australia continues to do some work in their ERP systems, based on the work of GovERP,” he said.

“The Secretary’s Data and Digital Committee really did some great work to assign it to our chief operating officer’s group.

“The DTA and [the Department of] Finance have special roles; the DTA has been looking at the technical capabilities that are necessary for it, the architectural alignment around the use of ERP, as well as the two [procurement]  panels.

Fechner added that the DTA and Finance are currently working with smaller agencies to determine what “their future ERP status is”.

“We are conscious of the sequencing,” he said. “SAP ECC support ends in 2027. We have extended that out to support it until 2030.

“But there are still a large number of entities that need to make fairly significant changes to their back office systems in the next few years.”