Insight 7: Invest in prevention and early intervention

Investment in prevention and early intervention is the most fiscally responsible way to improve the wellbeing of individuals and communities.

Investment in prevention and early intervention initiatives can result in long-term savings across the system because the need for future human service supports are avoided.¹¹⁵

Increase investment in preventative health care to improve wellbeing and avoid future costs

Preventive health involves taking measures to keep people healthy and well and to avoid the onset of illness, disease or injury. Preventative measures that reduce the incidence and severity of chronic health conditions lead to substantial long-term savings for the health system as chronic conditions are expensive to treat. Investment in cholesterol checks and blood pressure management can prevent heart attacks and the need for emergency care.

It is estimated that 38% of the disease burden for all Australians and 49% for Aboriginal and Torres Strait Islander people, could be prevented through prevention and early interventions by reducing health risk factors such as obesity, physical inactivity, dietary risks, and alcohol, tobacco and other drug use.¹¹⁶

Studies show that every dollar invested in preventive health saves an estimated $14.30 in healthcare and other costs.¹¹⁷ In addition to the avoided costs of future health services, healthier individuals are more productive, with lower workplace absenteeism and the ability to remain in the labour force for longer.

In 2022, Australia spent 3% of total health funding on prevention, which was equivalent to the OECD average but less than countries like Canada where 7% was spent on prevention.¹¹⁸ Building on the successes under the National Preventative Health Strategy 2021-2030, further efforts are proposed over the long-term to rebalance spending towards prevention.

Workshop attendees emphasised the importance of a social contract between government and communities, where it feels like the government is investing in their long-term wellbeing through preventative or early interventions. For example, Australia’s Long Term National Health Plan demonstrates a long-term commitment to strengthen community health services integrated in Australia’s primary healthcare system.

Shift investment in human services to the achievement of long-term outcomes

Refocusing decision-making and funding of services towards the achievement of outcomes over the long-term will maximise the cost-benefits of human services investment. The Budget Process Operational Rules (BPORs) in Australia can act as an inhibitor to investments with longer-term socioeconomic benefits, such as prevention and early intervention initiatives. A long-term outlook will likely shift investment towards prevention and early intervention initiatives which will have the highest impact.

Experts provided a strong evidence-base that early intervention services achieve outcomes for Australians and cost-savings over the long-term. Community members agreed that providing early support for people is critical for better outcomes. There is also evidence for broader economic benefits through productivity gains.¹¹⁹ The Productivity Commission estimates absenteeism and lost productivity due to mental health challenges cost the Australian economy up to $17 billion every year.¹²⁰ Early interventions that aim to help individuals proactively manage mental health challenges before reaching crisis point are also contributing to productivity gains for the economy.

There is the opportunity for government to consider collaborative funding proposals across agency silos for initiatives that deliver multiple outcomes. As an example, see the case study below on Victoria’s Early Intervention Investment Framework.

Another opportunity is to improve information and access to prevention and early intervention services at key ‘stages in life’ and for people who are at risk based on known factors. This could extend to proactive outreach to people when they become eligible for a particular service. The provision of information and outreach could be from a trusted source such as a family doctor or the local Aboriginal Medical Service and personalised to the circumstances and needs of the individual.

Insight 8: Harness the power of technology and data

Emerging technologies, including artificial intelligence (AI) and big data hold the promise of substantial productivity benefits for government and human improved service provision for citizens.¹²²

This section will explore some of these opportunities and case studies for how they have been piloted in Australia and overseas, as well as noting some of the risks and other considerations involved in their implementation.

Maximise the benefits of artificial intelligence for human services

Governments and service providers in Australia and around the world are already leveraging artificial intelligence (AI) in innovative ways with many cost-benefits.

AI reduces the administrative burden. AI offers opportunities to streamline or eliminate routine administrative tasks including document review and automated customer service enquiries.¹²³ This can allow human services staff to refocus their efforts away from administration to client engagement and problem resolution that requires human judgement, oversight and empathy. For example, in intake interviews, AI-powered transcription software can take notes on and automatically summarise client history, enabling case officers to focus on eliciting information and framing sensitive questions appropriately. A similar system called Minute is already being used in the UK to summarise social worker debriefs with their supervisors.¹²⁴

AI reduces human error. Automation can reduce the likelihood of human error whether by service providers or clients and make bureaucratic processes easier. This can include by pre-filling known details in forms, or using predictive analytics to flag details that appear inaccurate for human review.¹²⁵ This could also help identify abnormal patterns that could indicate corruption or fraud risks.¹²⁶

AI helps staff make more accurate decisions. Leveraging advances in natural language processing technologies, governments are also developing virtual assistants for both back-of-house and customer-facing applications. For instance, the Caddy co-pilot helps customer service advisors at the UK frontline charity Citizens Advice quickly find and share information about government services with people who have come to Citizens Advice for help.¹²⁷ In Australia, Services Australia uses an in-house virtual assistant called Roxy to direct staff to relevant resources to help resolve or escalate system processing issues.¹²⁸

AI supports better predictive analytics. The pattern recognition capabilities of LLMs can be leveraged to analyse vast administrative datasets to uncover previously undetected user needs that could the design of future service provision.¹²⁹ For example, the UK is establishing a National Data Library which uses NHS patient data to develop AI models that predict various kinds of diseases. In Australia, CSIRO has developed the Patient Admission and Prediction Tool (PAPT) to predict hospitals’ daily patient load, including how many are likely to be admitted or discharged.¹³⁰

AI supports personalised services. Generative AI chatbots could also provide a more seamless, personalised experience for users. A chatbot that recalls context wouldn’t require service users to re-tell their story multiple times to many different case officers or providers.

Understand and manage the risks of AI

The scope for productivity gains through application of AI to human services is limited by the labour-intensive nature of much human service provision.¹³¹ However, as human service provision forms an increasing proportion of the Australian economy and workforce, it will be essential to incorporate AI and other emerging technologies in some way.

Machine learning algorithms are fundamentally opaque, and it remains technically difficult to explain how inputs were associated with a particular output.¹³² This means that AI systems cannot be held responsible or accountable for decisions or outcomes. This issue of lack of accountability is most salient when used to make decisions about individual citizens’ compliance. The Robodebt case, while not an example of artificial intelligence, shows the risks when governments employ automated decision-making without human oversight.

AI outputs are also probabilistic, meaning cases may not be assessed on their specific legal merits.¹³³ When AI systems are used in government decision-making with ramifications for citizens’ lives, there must be ways to ensure accountability and recourse to ensure just outcomes. The EU General Data Protection Regulation (GDPR) provides citizens with the right to know, the right to an explanation, and the right to have a decision made by a human.¹³⁴

One area where AI is likely to be most useful for policymakers is the capacity of machine learning to analyse data and make inferences based on that data. However, it is important to be alert to the limitations of AI for analysis. While computers can conduct fine-grained descriptive analysis of data at a scale and speed that is beyond human capabilities, these are correlations of observational data only and on their own do not form a basis for causal inference. Outputs from AI must be reviewed in conjunction with other impact evaluation methods, including randomised controlled trials.

As government considers the opportunities for expanding the role of AI in human service delivery, it may be worth considering how to ensure technical systems reflect desired social values. Australia’s AI ethics principles emphasise a focus on fairness, privacy, transparency and explainability, reliability and safety, contestability and accountability.¹³⁵ Some options that have been suggested include supporting the development of enterprise generative AI systems that can hold sensitive personal information in a secure way on local Australian servers rather than being incorporated into global LLMs.

Maximise utilisation of administrative datasets to inform human services design and delivery

The expansion of large administrative datasets is also presenting new opportunities for human services policymakers, with high-quality, linked, de-identified person-level data enabling powerful modelling of social outcomes both at individual and population level. The ABS established the Multi-Agency Data Integration Project (MADIP) in 2015, which has since been renamed the Person-Level Integrated Data Asset (PLIDA).Leveraging administrative datasets can provide a richer picture for how social welfare needs are likely to change over time and within Australian communities.¹³⁶

AI for human services and of large administrative data are two trends which are both growing and interdependent—the quality of generative AI models is a function of the data it is trained on. Access to high-quality, population-level datasets often held by government is an area of interest for generative AI companies. For instance, foundation model company OpenAI signed a strategic partnership with the UK government in 2025, partly with a vision to transform public services.¹³⁷

Advancements in data-driven social science techniques, as applied to administrative datasets, are minimising the requirement to rely on survey data for social research, instead enabling tracing individuals’ life-outcomes throughout complex human services systems. In New Zealand, this investment approach modelling approach has been used across government to shape human services policy and strategy. Introduced in 2011, the New Zealand government started conducting actuarial analyses of integrated government data to identify points for early intervention to reduce future social welfare liabilities, based on Integrated Data Infrastructure established by Statistics New Zealand.¹³⁸ This data asset is longitudinal and covers all New Zealand residents and integrates data from non-government organisations. New Zealand found that human services costs are unevenly distributed across the population, concentrated in a long tail of those with highest support needs. The findings demonstrate that universal human services are working for over 80% of the population and that significant cost savings can be made by focusing on the most vulnerable cohort.

Progress is being made in Australia with the Department of Health, Disability and Ageing’s (DHDA) HOME application that enables system exploration and predictive modelling of the health system. HOME brings together PLIDA with other national-level and state and territory person-level data assets. This enables an understanding of human services costs, and outcomes, down to the individual level, enabling an integrated explanation of interactions between the health, disability and aged care systems. HOME also has the capability to combine person-level datasets with simulation-based modelling to see how different scenarios, including changing demographics and fiscal pressures, might result in different system outcomes.

A national digital map of human services could be created to draw together multi-jurisdictional and multi-domain human services data and provide insights to drive decision-making. High quality human service maps can overlay community need with existing human services in a location and identify gaps and duplications in servicing. While community representatives working in rural and remote areas emphasised underspending as an issue, several service provider organisations raised the question of whether the larger problem is how strategically (or not) funding is used by government. Service providers suggested there is sufficient investment in the sector, but government does not always allocate it in the most effective way to meet communities’ needs.

Understanding and managing the risks and limitations of big data

Linked data is still a relatively new concept and social licence is still being built for it in the Australian community. Managing citizen concerns over data protection and privacy, especially when concerning sensitive health records is critical. Data linkage must be done in a privacy-preserving manner which anonymises data to the full extent possible when shared. Privacy-preserving techniques are rapidly becoming more sophisticated – in order to continue to build social licence for linked data, governments may use synthetic data for research and analysis purposes, which could replicate patterns in population-level data but would not reflect any individual’s personal data therefore preserving personal privacy.¹³⁹ In Australia, the Priority Investment Approach is an example where a synthetic longitudinal model of data has been used to model how Australians are expected to use social security over their lifetime.¹⁴⁰

One challenge is to increase policymakers’ understanding of what linked administrative data can and cannot tell us about service users’ needs. While linked administrative data assets offer significant opportunities, itis important to note they are skewed in distinctive ways that make them unrepresentative of the general population. For instance, PLIDA is likely to have a much richer data on an individual who is a net recipient of government benefits than of someone who is not. Linked data must be weighted or otherwise compared with census or representative population samples to ensure it accurately reflects the population of interest. Data quality may also be poor for other reasons, for instance missing data, time lags in data availability or challenges in analysing large amounts of unstructured data.

There is an opportunity to further improve the quality of this data and build a stronger data sharing ecosystem between all sectors. The Data Availability and Transparency Act 2022 has led to the implementation of a scheme to authorise and regulate access to Australian Government data. While a key policy issue for the current review of the DATA scheme is whether the private sector and non-government entities can access this public-sector data, there may be opportunities to facilitate sharing in the other direction in a secure, privacy-preserving way, in order to improve the quality of data, drive innovation and improve outcomes for Australians.¹⁴¹

Insight 9: Reimagine the role of government

A significant opportunity for the future of human services is to reimagine the role of government as stewards of the system.

This involves a shift for government. From market stewards commissioning services, managing contracts and ensuring compliance, to system stewards guiding a complex service delivery system towards the common goal of improving wellbeing outcomes for Australians. It involves working in partnership with trusted community providers, joining up and integrating service delivery and making the long-term wellbeing of Australian communities central to budget decision-making.

Move from market stewardship to system stewardship

System stewardship is a concept that brings together systems thinking and public administration to reframe the role of Government. UNSW explains how system stewardship involves supporting cooperation among multiple stakeholders to achieve shared long-term wellbeing outcomes for Australians.¹⁴²

This involves a fundamental policy shift from government being stewards of the market, to Government being stewards of the system. Moving from commissioning services, to guiding a complex service delivery system towards a common goal. Moving from outsourcing by default to only outsourcing when it will deliver the best outcome. Moving from managing contracts and ensuring compliance, to managing partnerships and coordinating across multiple service delivery organisations and trusted community providers.

Successful system stewardship involves building a common vision, sharing learnings, coordinating efforts, testing solutions, responding to local conditions, prioritising constrained resources and measuring wellbeing outcomes. This is in addition to government’s direct role in delivering human services, setting rules and ensuring compliance. It also requires the management of an increased level of risk. A case study in the Australian aged care sector demonstrates how system stewardship at the regional level can improve responsiveness, enable information flow, build capability of service providers and increase the focus on wellbeing outcomes.

We will know the system is set up to improve the wellbeing of people experiencing severe and multiple disadvantage, when we meet the nine criteria in Table 4.¹⁴⁶

Table 4: Systems that are effective in responding to severe and multiple disadvantage meet nine criteria.¹⁴⁷

Make the long-term wellbeing of Australians central to budget decision-making

Embedding wellbeing into budget processes will support Government to prioritise and evaluate public investment. Australia’s Measuring What Matters framework provides a national architecture for doing this, defining five interconnected wellbeing themes – healthy, secure, sustainable, cohesive and prosperous – and tracking progress across 50 indicators updated annually through the ABS dashboard.¹⁴⁸ These updates highlight both progress (for example improvements in life expectancy) and challenges (such as declining perceptions of safety). This wellbeing data can support decisions about future resource allocation, enabling decision makers to identify which areas are improving and which require strategic investment.

Victoria’s Early Intervention Investment Framework is a strong example of how budget bids can be assessed against wellbeing impact and avoided costs.