It is a truth universally acknowledged that there is huge potential for public services to be transformed by artificial intelligence (AI). Longevity is rapidly increasing. We are living longer and healthier lives and the population growth is slowing down. The working age population is shrinking and with that comes a reduced capacity to deliver public services. At the same time the demands and needs for public services are increasing and the pressure on efficiency of public services has never been larger. We need to do more with less.
AI may be able to improve efficiency, reduce costs, and even deliver new services also to groups who are currently unserved. Health apps, for example, are increasingly used by people who are not registered with primary care providers, and who therefore have no other access to healthcare.
Who is driving transformation?
There is a wide range of stakeholders involved and affected by the increased use of data, analytics and AI in public service delivery. These include:
Service providers keen to maximize efficiency and resource use
Even in publicly funded systems, competition is creeping in around the world, and resource pressures are ever present. Public service providers from education through health to police are ambitious to serve new groups, provide new or higher quality services, or make resources stretch further and serve more people for the same money. This adds up to an enthusiasm for new technology that could help them improve service delivery. Examples include Moorfields Eye Hospital’s partnership with Google on diagnostics, and systems being used in North Carolina to reduce the use of over-stressed police officers. Additionally the Royal Society recently published a very comprehensive infographic illustrating above.
Budget holders at national and local level, wishing to improve resource use
The level of government debt is rising in many Western countries to unheard-of levels. These governments have huge incentives to reduce the costs of the public sector, but so far, the only way has been to cut services. This is generally agreed to be unsustainable. Using AI offers potential for services to be delivered differently, and perhaps reduce expenditure in unexpected areas. Examples include prisons using analytics to match cell-mates whose interactions will reduce future reoffending rates, and the National Crime Agency trying to get ahead of criminals by using predictive analytics, rather than reacting after the event.
Citizens, wanting more efficient services, delivered more effectively
Perhaps the most important stakeholders are those who both use services and, ultimately, pay for them through their taxes. Citizens generally want high quality public services that are easy to use and accessible, but that are delivered well and efficiently. Since we have instant access to retail providers, we tend to expect the same of our public services. AI and data-driven services hold out the potential for this.
International institutions, trying to create coherent regulatory and ethical frameworks
European and other international institutions are taking a keen interest in AI. They have a role in establishing suitable pan-national ethical and regulatory frameworks for these developments. These include, for example, the new General Data Protection Regulation (GDPR), but also new frameworks for ethical use of AI, including transparency of decision-making.
Overcoming the challenges: measuring success
There are many challenges to exploiting AI in the delivery of public services. Perhaps not the least of these, in an increasingly target-driven environment, is how to measure progress and identify success.
It is easy to say that KPIs will be needed, and that they should encompass productivity, revenue and social issues. But what exactly should they be? Perhaps efficiency and reduced costs are key, since those are an important part of the business case for adopting AI. But at the same time, it is important to monitor service quality, to avoid a ‘race to the bottom’, and ensure that AI-delivered services really are at least as good as anything they replace. The key question is probably whether each project delivers on its aims, although that requires careful definition of those upfront, including a check on whether they are appropriate, and in line with user requirements.
It might also be helpful to use citizen and user satisfaction as a key measure of success. If AI is designed to improve service delivery, then why not ask whether users would recommend the AI-delivered service to others?
The importance of trust
User recommendations may be important in more than one way. The success of AI delivered services depends on whether citizens are prepared to engage with them, and that, ultimately, depends on the development of trust in these services. No service can succeed without users. Users and taxpayers will need to be sure, for example, that their personal data will be used appropriately and within guidelines, and kept safe. They will also need to be confident that the service will be at least as good as before. Public service providers will have to work to build this trust if AI is to deliver on its promise in public services.