How Generative AI Can Open New Frontiers in Public Service Delivery and What's Needed to Get There

ETHOS Issue 27, Forthcoming

Generative artificial intelligence (GenAI) is only the latest frontier in a suite of new technologies that will profoundly reshape the nature of work, citizen engagement and service delivery. These are early days: most public and private sector organisations globally are still planning and experimenting with ways to deploy it meaningfully. What is already abundantly clear is that GenAI is shaping up to be a catalyst for ground-breaking change across economies, markets, societies, and governments. It will significantly impact the nature of work across all industries, either through automation or augmentation.

BRACE FOR POSITIVE IMPACT

Accenture’s research across 22 countries and 19 industries suggests that the public sector will see 42% of its working hours impacted by GenAI.¹ There is much room for optimism that this impact will benefit the public sector in diverse ways, from reinventing organisational processes and redesigning workflows, to altering workforce dynamics and influencing regulatory environments.

For one, GenAI can be used to tap the full potential of data, to better illuminate citizens’ needs and assess policy outcomes. Co-pilots trained on Large Language Models (LLMs) can also automate data retrieval, helping civil servants draw more efficiently from rich troves of data across multiple channels to prepare reports in significantly less time.

These AI tools can also optimise language-based work processes, such as by automating chatbot reminders and case progress updates to clients.

OFF TO A PROMISING START

The Singapore Public Service is no stranger to leveraging artificial intelligence to make services more efficient, seamless, and integrated for citizens, through initiatives such as the Digital Government Blueprint and the National AI Strategy. It is this long-running strategic focus on harnessing advanced technologies to transform the public and private sectors that have won Singapore the top spot in Asia’s Artificial Intelligence (AI) readiness rankings.²

Agencies like GovTech are already reaping results from AI pilots. Its One Service Chatbot automates the extraction of case details, predicts the category of cases, and then routes them to relevant agencies—saving about 2,000 man-hours since it began operations in 2021.

There is potential to further enhance the One Service Chatbot with functions already possible with ChatGPT 4, so it might evaluate both text and photos to address citizens’ concerns. For instance, users could submit a picture of a cluttered HDB corridor, and the Chatbot could then automatically generate a case for resolving this fire hazard.

Soon we might also see government agencies harness AI-powered translation and voice generation—for instance, to boost the diversity and reach of public service education videos for Singapore’s multilingual population.

These are just some examples of the immediate benefits that the public sector might reap from GenAI approaches. But for GenAIs to serve as valuable co-pilots, their LLM models must be enhanced with agency-specific datasets, employing techniques like fine-tuning and instruction-led tuning. LLMs are a type of foundation model, designed to understand, generate and interact with human language. Such models need vast amounts of curated sample data to learn effectively.

This scale of data requires a modern enterprise platform to host a trusted, reusable set of data products. Such enterprise-wide platforms liberate data from organisational silos by securely housing the data, along with enterprise-grade analytics and AI tools, in cloud-based warehouses. These data warehouses can be made easily accessible for use across organisations or through different federated computing techniques, such as a data mesh.

As the Singapore Public Service develops a robust core digital and cloud-based infrastructure of its own and becomes more data-mature, it will be able to push the boundaries of what AI co-pilots can achieve. For example, it could employ active learning techniques to train co-pilots on agency-relevant data in real-time for accurate, timely decisions at scale.

The true value of foundation models becomes evident when these models are contextualised to provide accurate and nuanced answers to user queries. Contextualisation may involve fine-tuning the foundation model on organisational data or using a more scalable approach like Retrieval Augmented Generation (RAG). RAG enables the GenAI system to "search" against institutional data and embed this into the system's results, offering richer and more contextualised outcomes.

For instance, a RAG-enabled LLM for public housing in Singapore might provide specific and tailored answers to home buyers navigating the range of housing grants and policies. Similarly, businesses could find it easier to interact with a chatbot using a RAG-enabled LLM for taxation policy, which may otherwise be technically challenging for the typical business owner to comprehend.

BOOSTING GenAI UPTAKE

The Singapore Government has been an early mover in piloting and accelerating GenAI solutions. One notable initiative is Project MindForge, for which the Monetary Authority of Singapore is collaborating with partners like Accenture to unravel the complexities of GenAI and draft a risk framework guiding its responsible integration into the financial sector.

For governments to harness GenAI effectively, they will need to do much more than implement modern data platforms and new technologies. They will also need a governmental workforce equipped with the Technology Quotient (TQ) to drive GenAI at scale. Nurturing TQ starts with encouraging workers at all levels to learn about this emerging technology, how it might be applied, why it matters, and how it works with other technologies.

Everyone in an organisation, from senior decision-makers to front-line staff—regardless of skill level—should develop a strong appreciation of GenAI’s capabilities so that they would be motivated to adopt it. Only then can work and the workforce be truly transformed to tap the full potential of AI.

Public sector leaders must take the lead in driving this workforce transformation by contemplating and addressing these key questions: How should we approach workforce planning to effectively harness GenAI? How can we effectively reskill and train our personnel to adapt to this new technology? What cultural shift should we anticipate?

RESHAPING WORKFLOWS TO EMPOWER WORKFORCE

But before leaders race ahead to reskill their workforce, they need to first immerse themselves in the technology and embed continual learning into their workflows. Once they get a clearer view of where in the entire workflow might GenAI be most impactful, they will be able to identify opportunities to reallocate work in ways that free employees up to perform at their best. This includes aligning GenAI technology with agency goals for greater efficiency, restructuring or collapsing organisational silos in a lasting, meaningful way.

The ultimate goal of workflow redesign is not just to embed GenAI technology into the organisation, but to embed the value AI brings across every work process. Be it enhancing service quality, improving policy outcomes, or providing the robust support staff will need to raise their TQ and productivity, reinventing work processes around AI must ultimately empower an organisation to become as agile and dynamic as the technology they harness.

A PEOPLE-FIRST APPROACH

GenAI’s effectiveness hinges on human input to drive quality outputs. Technologies that are “human by design”, including spatial computing, body-sensing technologies, and AI-empowered agents that work on behalf of humans in an interconnected ecosystem, will further transform human-machine interactions and enable humans to accomplish things once thought impossible.

AI-infused processes will affect practically every role in the public sector. Accenture’s analysis across 22 job categories such as administrative support, management, healthcare practice, and social service has revealed that software utilising GenAI and Traditional AI will impact each category by as much as 63 percent of a workday.³

This shift calls for public officers to pivot away from repetitive or time-consuming tasks that AI can quickly resolve, towards more cognitively complex and judgement-based tasks. They must learn how to effectively use GenAI as their “workmate” to meet citizens’ needs.

This may be daunting for some workers, leading them to resist the changes brought about by GenAI. Some may feel they lack the skills and confidence to fully benefit from using the technology.

Others, feeling the pressure to rapidly acquire new expertise in their redesigned roles to avoid being displaced by GenAI, may view it as a threat rather than a tool.

To boost employees’ confidence in GenAI, leaders need to take a people-first approach. Success with GenAI demands as much attention to people and training as it does to technology. Therefore, the public sector should significantly increase investment in talent to tackle two distinct challenges: creating AI and using AI.

Creating an AI-empowered public service will certainly entail developing expertise in technical domains such as AI engineering and enterprise architecture. But beyond building technical proficiency, leaders must also train personnel across the organisation to operate effectively within AI-integrated processes. While AI is unlikely to bring about the full automation of jobs, many tasks are set to evolve. The competitive edge will come from investing in task-oriented training and collaborative work between humans and machines.

This training should encompass not just technical know-how, but human-centric skills as well. Higher-level cognitive thinking and empathetic problem-solving will become increasingly important for public officers looking to effectively apply AI to serving citizens’ needs in more intuitive and relevant ways.

AN “AI-POSITIVE” CULTURE

Accenture’s research shows that 97% of public service employees worldwide want to acquire GenAI skills. Indeed, they are more optimistic about GenAI and its impacts than their peers in other industries. However, only 7% of public service organisations are reskilling their people at scale. This is a gap leaders must address urgently.

Individuals may be initially motivated to experiment with GenAI, but their progress may be hampered without the consistent practice of using AI on a daily basis, as old work habits may die hard. Leaders will need to jumpstart a significant cultural shift.

For the public sector, this will entail developing a culture that embraces innovation and encourages employees to maximise GenAI’s potential for the public good. This will also be a culture where people feel they have greater autonomy and fulfilment in their careers, being empowered by AI to move beyond executing tasks to shaping their own work and organisational workflows.

To drive this change, leaders need the vision and boldness to create environments where people feel inspired and motivated to keep adapting work, learning, and changing at a whole new level made possible by AI.

USING AI RESPONSIBLY

At the heart of an AI-positive culture is a responsible AI mindset and set of guiding principles to build systems which ensure trust, transparency, and fairness. This also demands a modern data platform with data security measures in place.

Responsible AI is a critical capability to assess the potential risks of AI systems, and it becomes even more crucial for GenAI systems. Embedding Responsible AI in an organisation means the approach needs to be considered at each design stage and incorporated into processes and systems throughout the public sector.

This is especially important because the public sector often serves the most vulnerable members of society, and decisions made by agencies assisted by chatbots can often be life-changing. Initiatives such as Singapore’s Model AI Governance Framework for Generative AI (MGF for GenAI)⁴ are laying foundations for a trusted ecosystem for developing these technologies. Globally, the World Economic Forum’s AI Governance Alliance is also bringing together over 260 organisations across industry, government, academic, and civil society to deliberate this issue. In 2023, the Presidio Recommendations on Responsible Generative AI offered 30 guidelines for the ethical use of generative AI, emphasising responsible innovation, international cooperation, and societal benefits.⁵

To pave the way, leaders must educate their staff about the security, legal, and ethical considerations of AI, particularly the issue of transparency about where and how GenAI is being used to make policy decisions and deliver services. Leaders should also foster an environment of transparency while educating their staff about how to effectively handle the risks of GenAI. Armed with greater awareness and knowledge about managing GenAI’s benefits and risks, public service staff will have greater confidence in the use of responsible AI. This in turn will boost public trust in GenAI.

EMBEDDING AI ACROSS THE PUBLIC SECTOR

At Accenture, we understand the needs and issues faced by leaders in integrating GenAI into their workforce and workflows. About 600,000 of our people across all levels are already trained in data and AI fundamentals. This goes a long way in embedding GenAI into how we deliver our services and how we operate.

The journey towards an AI-powered public sector will call for strong leadership to reinvent work, reshape the workforce, and foster collaboration between civil servants, technologists, and the public to build an ecosystem of responsible AI. By leading and learning in new, continual ways, Singapore’s Public Service can enable a more accessible, citizen-centric, thoughtful future in the Age of AI.