Most public sector organisations are already exploring or actively working on generative artificial intelligence (GenAI), and up to 90 per cent are planning to explore, pilot or implement agentic AI in the next couple of years.
This finding was based on Capgemini’s latest report surveying executives from 350 public sector organisations around the world.
While GenAI excels at creating new content, agentic AI focuses on making autonomous decisions and taking actions to achieve goals.
“With GenAI, human-in-the-loop (HITL) allows human interaction with AI systems at various stages. You see the outcomes, then you determine and decide what actions to take,” said Capgemini’s VP & Head of Insights & Data Global Business Line, APAC-ME, Dr Kirti Jain.
He added that agentic AI is transforming the next phase of AI: “This means that instead of waiting for a human to take the next step to trigger the action, the machine can run itself through the last mile.”
While there is an interest to explore more advanced AI use cases, data trust, governance and security remain the biggest challenge for these organisations, with 64 per cent of them expressing concerns about data sovereignty.
More concerningly, less than 25 per cent of organisations have the data needed to train the AI models.
As the Singapore public sector begins to explore agentic AI, Dr Jain made the case for agencies to put data governance considerations at the onset before designing AI to ensure consistency, accuracy and quality in data and decision-making.
Instead of having to rebuild it later, data governance should be an integral part of design thinking.
To subscribe to the GovInsider bulletin, click here.
Data governance-by-design
Dr Jain emphasised on the importance of proactive data governance, contrary to reactive incident response, as the critical foundation of any successful data-driven organisation.
He elaborated the scope of governance management to all the steps of data lifecycle – acquiring, managing, and utilising data.
For public sector agencies, a flawed AI-generated insight could lead to decisions with significant negative impacts for citizen services on a large section of people, Dr Jain said.
He shared three areas that Capgemini supports public sector agencies when it came to data governance management.
The first was helping agencies build a secure data infrastructure.
This was followed by ensuring data usability by improving the data quality, addressing privacy concerns around data use, and implementing governance to manage data access.
Finally, it was in building solutions that enabled more efficient citizen service delivery and data-driven decision-making.
The same consensus was shared by Australia’s Environment Protection Authority Victoria (EPA)’s Chief Technology Officer, Abhijit Gupta, as he highlighted essential steps for public sector organisations to advance their data and AI journey.
“Data security, privacy, and timely data activation are all critical for public sector organisations. It is important that data is visible and usable for business purposes,” he said.
Scaling agentic AI
Be it a human or a machine, Dr Jain pointed out that users still expect accuracy and consistency in the eventual outcome.
“In an ideal world, data governance should be agnostic whether it was with humans or machines,” he said, adding that there is need for even more agility in governance as technology evolves and things change more rapidly.
And thus, with a robust foundation in data governance, only small, incremental changes need to be made to scale agentic AI, which can automate end-to-end decision-making.
“Our focus is always on proactive governance of data assets for any enterprise – this is principle of active data governance. Once this foundation is established, you can build as many AI initiatives as possible on top of it,” he noted.
Capgemini earlier published a data maturity model that provides a framework for organisations to assess their capabilities in gathering, processing and utilising data.
Singapore well-positioned to accelerate emerging tech
Dr Jain highlighted Singapore’s experimental and collaborative approach that puts the country in a position to accelerate the implementation of new technologies, including agentic AI.
He cited the AI Trailblazers initiative as an example, which is a government-led initiative, partnering with the private sector, to identify and address 100 and more real-world challenges from GenAI solutions.
Given the rapid evolution of technology, experimentation and feedback are crucial for continuous improvements, he added.
By embracing a “fail fast, learn fast” mentality, the Singapore government has quickly identified the successful solutions, scale them up, and adapt to the evolving circumstances.
Additionally, Singapore takes a collaborative approach in defining the guidance and guardrails needed to govern AI and GenAI initiatives, be it sustainable AI or AI governance.
Recently, Singapore’s Minister of Digital, Josephine Teo, highlighted Singapore’s commitment to sharing best practices with the international community at a recent conference.
“We will learn, and we will make mistakes. The mistakes that we make, not everybody has to go through them”, she said.
Dr Jain added: “This is why Singapore is moving so rapidly in the advancement of not only technology, but how technology is used for betterment of citizens and the public sector”.