Transforming public services: the intersection of AI, data and purpose

When government agencies are purpose-led in leveraging data, they can boost public service value and community confidence.

In brief:

• Advances in AI and data have tremendous potential to generate actionable insights that will revolutionise public services and improve productivity.
• Government requires thoughtful processes, guardrails, and practices to ensure AI augments human judgement to deliver in the public’s best interests.
• Being purpose-led will ensure agencies are deliberate in how they use the right data from trusted sources, for the right reasons.
  

The combination of generative artificial intelligence (GenAI) and data has a powerful potential to support local, state and federal governments in both foreseeing and understanding the diverse needs of their citizens, businesses and the community to deliver better social and economic outcomes.

The volume and speed at which data is being generated is already enormous and its growth rate is only accelerating. Most of the world's current data holdings were created in just the last two years. Government agencies are only just beginning to unmask the capabilities of GenAI to use this massive resource. Yet, already, concerns are being raised about the ethical use of data and AI.

High-profile case studies illustrate the point that even well-intentioned technology solutions can result in poor community outcomes. To navigate the vast sea of information responsibly and effectively, using tools whose potential is barely understood, data leaders and policy makers need an overarching strategy that ensures data and AI are used for good.

To support this, the Australian Government has proposed 10 mandatory guardrails for high-risk AI, as well as a Voluntary AI Safety Standard with practical guidance on responsible AI implementation. The guardrails are for high-risk AI settings, taking into account the potential for AI systems to cause harm to individuals, groups and society as a whole, including from a human rights, health and safety, and legal perspective.

This article broadens the lens, looking at all situations where agencies employ AI and data. In all cases, we believe:

Purpose is the ultimate guardrail

Data and AI are powerful enablers, but they are not the end goal. No matter how powerful and insightful they may be, their use must always be driven by “purpose”. What benefit is the agency hoping to achieve for citizens or businesses?

This purpose should influence everything from data collection and analysis to decision-making, policy implementation and measurement.

This focus on purpose must flow down into individual AI-related projects. EY teams have been working across Asia-Pacific on AI and data implementations to support various aspects of digital trade. For example, a major port is now using AI models to detect anomalies in the live sensor data being thrown off by cranes. The models flag otherwise unknown signs of failure and recommend maintenance actions to minimise downtime. In the same port, AI models are also accurately predicting ship arrivals so workers can be rostered on only when needed – delivering annual savings in millions for the port operator. In a customs agency, machine learning models now identify import permits with a higher probability of containing contraband – improving the rate of success for search teams.

In all three examples, our purpose was to improve productivity. With that as our guiding principle, we had to deliberately calibrate the AI models for a low rate of false alarms. Because swamping teams with multiple alarms would have created time-wasting work – defeating the purpose.

When facilitating conversations about how to architect a federated data model, the key is to keep drawing participants back to purpose. This means asking the human-centric questions:

• Will that create additional costs or administrative burdens for businesses or government?
• What’s the minimum set of data required to achieve our purpose?

This is how to ensure that efforts to get better outcomes don’t result in unintended consequences for the people and businesses agencies serve.

How can data stop homelessness before it starts?

Each year, governments expend resources on dealing with the impact of homelessness after it reaches a crisis point. But the problem continues to grow. Now, data and AI are helping to augment the insights and experience of case workers to help agencies better understand the predictors and signals of when people are mostly likely to become vulnerable.

By integrating, matching and merging data from disparate organisations, it is possible to form a complete picture of a vulnerable person. Then AI can be used to apply predictive analytics to make earlier interventions in those cases where it is possible to prevent a crisis from occurring.

This was the goal of the UK Borough of Maidstone which, in 2018, experienced a 60% rise in the impact of homelessness. EY teams were asked to design a new, preventative Homelessness and Rough Sleeper strategy. Together with a data and analytics partner, we designed and built ‘OneView’ – a data and analytics tool that brings together data from various sources to identify those at risk of future homelessness.

The thinking behind OneView is that, while homelessness is complex – often rooted in multiple and cumulative disadvantages – it can also be caused by a simple change in circumstance requiring a temporary solution. When that is the case, issues can be resolved comparatively quickly, easily and cheaply to prevent families from ending up on the street. Then specialist services and resources can be dedicated to the more complex cases and those most in need.

Three questions to help link data to purpose

It is easy to become lost in the intellectual pursuit of technical possibilities, rather than constantly coming back to the “Why?” behind a data-led initiative. The following questions will help agency data teams make the connection between their work and purpose:

1. How do we frame the scope of any data-based activity with its purpose (and people) at the centre:

• “How will this support those who are vulnerable or at risk of harm?”
• “How will it drive business value for our organisation?”
• “How will it build community and business confidence?”

2. How can we ensure investments in data and AI align with our agency's purpose and values?

3. How can we foster a culture of purpose-led data use within our agency?