For government, being able to streamline or automate high frequency, low risk decisions is at a critical stage. Faced with a flood of information, government departments, public sector leaders and executive groups are exploring new decision-making techniques and tools.
Advanced analytics is opening powerful opportunities to personalise government services and focus attention on the key drivers of social problems and their potential solutions. The shift from a mass government approach to precision delivery is enabling public agencies to target resources where they are most needed, while standardising processes for the majority.
While policy makers have long sought to target their efforts, artificial intelligence (AI), combined with a powerful suite of visualisation and other tools, is enabling decision makers to better triage decision-making, thereby enabling government agencies to automate manual processes and freeing them to focus on high value/high return issues.
This was the key message delivered during a recent seminar in Canberra hosted by The Mandarin and supported by SAS.
The seminar heard from the former Secretary of the Department of Energy and Environment, Gordon de Brouwer; the former Director-General of IP Australia Patricia Kelly; and Deputy Commissioner Smarter Data with the ATO, Marek Rucinski.
Observing that analytics has a long heritage in policy making, de Brouwer said much of the strong economic analysis that traditionally underpins federal government decisions was based on similar systems thinking.
He said advanced analytics had been used around environmental decision-making and he saw significant benefits in removing some of the bias that can creep into traditional human decision-making.
He also noted that analytics enabled agencies and departments to identify the leading indicators that will signal when a policy or project is going off track – allowing for pre-emptive intervention.
These approaches improve the quality of outcomes and reduce the cost of services by automating currently manual bureaucratic processes.
Patricia Kelly said IP Australia had taken a strategic decision to embrace analytics, focussing on applications that help customers and staff. IP Australia was one of the first Commonwealth agencies to trial AI to automate existing processes.
Marek Rucinski oversees one of the biggest analytic shops in Canberra, involving over 500 people. Joining the ATO from Accenture earlier this year, he has moved to consolidate the ATO’s analytic play into a hub and spoke model, keeping close to the business units and aligning analytic initiatives to the business objectives of the ATO’s main work programs.
The ATO has been a leader in analytics, applying robotic process automation to enable prefilling of tax returns; applying AI applications to observe outliers; and applying AI to its vast precedent library of decisions for better and more consistent decision-making.
Departments are also being pushed to report on and manage outcomes; and look for easy-to-understand, reliable reports, focussed on key agency performance metrics and the allocation of precious resources.
For federal government this is being driven by the rollout of the governance, performance and accountability regime to support the PGPA Act. This regime highlights the need for agencies to develop and manage real-time performance and outcome metrics.
It was observed that when applied across agencies this would enable a calibrated approach to risk management, enabling agencies that were performing within agreed parameters to need less oversight and constraints. This comes as the APS review considers how to improve performance and outcomes across the Australian government.
Improving service delivery requires finding process efficiencies
Addressing a demand that already exists, the seminar heard that a key role of AI in modern and future government is to deliver faster and better services.
Rucinski described automation and AI as a by-product of scaling up and industrialisation. If a system or process can be described, it is programmable and therefore can be automated. By automating time consuming manual tasks, government can help staff to increase efficiency and focus on more complex problems. Data input is an example of a process that can be automated by pre-filling.
Kelly explained that over the past four years, IP Australia has seen customer transactions move from 12 percent online to 99 percent and that as a result, customer service and satisfaction had improved.
Artificial intelligence doesn’t replace human labour – it augments it
The panel observed that AI augments human decision-making by tapping the information held in case files as well as enabling agencies to reduce the cost of services by automating currently manual bureaucratic processes.
While AI supports the automation of manual labour roles, it does not remove the requirement for human labour. It just shifts it to a new function. Kelly observed that AI frees people to focus on what we are best at: high level decisions that require creativity and empathy.
Rucinski described the role of people in the system as transitioning from ‘spooners’ to ‘solvers’.
“We humans are still the front line,” he explained.
de Brouwer observed that if the introduction of AI was led by staff – and not top down – it would increase chances of success.
“People are highly motivated,” he said. “If we see how it can improve outcomes, we will embrace it.”
Requiring less staff to support data creation and processing frees them up to validate, verify and interpret results – a skillset Ray Greenwood, Domain Lead with SAS, emphasises government will need to invest in.
Among the changing roles will be increased demand for data scientists and programmers. While these are currently and commonly found in back rooms of government agencies, Rucinski urged agencies to consider them as part of the future frontline of operations.
“Playing around is not good enough anymore, it’s time for AI to move from back room experimentation to frontline adoption” he said.
AI is easier than it sounds
Government shouldn’t shy away from starting their AI journey.
Despite different levels of AI maturity within the government sector, agencies can start their journey even if they have small innovation spaces or platforms that aren’t perfect. Greenwood explained it was important to put a stake in the ground on what was acceptable and get started.
“We are never going to get perfect decision-making engines,” he said. “By way of analogy, even though DNA testing has been shown to be imperfect as a means for identifying criminal suspects it still provides a massive improvement in guiding decision making in criminal cases and is therefore a powerful tool.”
Given that AI is a trend that will continue to grow, it makes sense to get on the front foot.
In getting started with AI, the speakers agreed it was important to focus on low hanging fruit, such as customer engagement – which were likely to be areas with highly repetitive tasks that consume human capital.
Rucinski said it was important to spend time engaging with stakeholders who will be using the solution at the end. Equally, it was important to bring the agency on the journey. A customer focus was central to this process.
But to realise the potential of AI, it was important to show viability early and move on to the next win.
“Keep developing new use cases and embrace the process,” Rucinski said.
Challenges remain in scaling up
“The technology is there and it is easy to get initiatives to test them,” Greenwood said. “The challenge is to get something that outlives the individuals involved.”
In developing solutions using AI, he urged agencies to engage in a conversation of how AI and associated solutions could take on a life of their own – and what structures would need to be put in place to enable them to scale and continually improve.
Often the focus was on the now, and not the future – despite the discussion recognising that AI is transforming government today.