Beyond automation: Agentic AI as the next generation of public service efficiency

Agentic AI is emerging as a transformative system that can reason, plan, remember, and act without requiring continuous human intervention.

These capabilities are increasingly being integrated into software systems to perform tasks on behalf of users, which increases efficiency and quality of services when deployed correctly.

“If I said to you ‘we’re rolling agentic AI out in your organisation tomorrow’, who thinks that your organisation would be ready for that adoption?” said the interweave.gov’s Co-Founder and Author Luke Cavanaugh.

He was moderating the session Agentic AI: The Next Leap in Public Sector Innovation, at GovInsider’s Festival of Innovation 2025 when he posed the question to the audience, to which no one raised their hands.

Cavanaugh continued the line of reasoning citing a survey conducted by Microsoft that when asked if agentic AI has a lot of opportunity, almost all the respondents would put their hands up in agreement.

While some organisations still remain hesitant about adopting AI agents, the technology is here to stay, with public officials recognising its potential for workplace revolution while navigating its adoption.

To make the best out of this technology, the speakers emphasised the importance of thinking strategically about organisational goals and capabilities to identify the best-fit areas for AI and automation, and shared practical insights to integrate this technology into existing workflows.

To subscribe to the GovInsider bulletin click here.

Working beside AI agents

In the public sector, there are various areas in which agentic AI can lend a hand to improve processes and enhance human capabilities.

One of the examples shared in the session was about deploying AI agents to work alongside facility managers in building management. This use case was shared by Shermaine Wong, Senior Manager for Smart District of Singapore’s JTC Corporation.

For example, if there is a problem with an air handling unit (AHU), both humans and AI can work together to identify the problem and act proactively toward a solution.

“The facility manager may draw on his years of experience to know that an overcooling may be due to an opening of the valves… an AI agent can also perform this reasoning,” she said, adding that once the reasoning happens, the agent can test out alternatives to determine the root of the problem and arrive to a fitting solution.

The example shows the agent’s ability to interact with the external environment and operate autonomously, helping facility managers to streamline processes and adopt more time-efficient solutions, Wong noted.

Throughout the session, the speakers highlighted the importance of maintaining a level of human intervention in certain processes while leveraging agentic AI, especially in the initial stages.

In the context of healthcare, having medical informatics specialists in the loop is crucial to ensure that the agents are performing correctly, said Singapore’s Ministry of Health (MOH), Health Analytics Division’s Director, Sutowo Wong.

“Human intervention is important because in healthcare, we are talking about human lives,” he said.

Still, healthcare is an exciting sector with different opportunities to test out AI agent designs in the view of the Director.

There could be a genomics data agent that decodes data from a biopsy sample to identify biomarkers, or a specialist agent that analyses data from doctors’ clinical notes using natural language processing and initiates workflows for future steps, MOH’s Wong shared.

Given the wide range of use cases in healthcare, deploying agentic AI in existing workflows would require clear guidelines and a transparent “chain of thought” to oversee how the agents operate in that context, he noted.

Thinking strategically

“It must have an alignment to the business outcomes and the goals. What is the agentic AI trying to do and what do you want it to do?” said Workato’s Solutions Consulting Director, Ken Ng.

Ng shared that adopting a strategic approach is essential for organisations that are interested in deploying agentic AI on a larger scale.

To ensure that the AI agent works effectively, the processes and targets must be clearly stated in such a way that the agents operate according to organisational objectives and expectations.

Moreover, Ng urged listeners not to overlook the importance of getting the basics right: connecting systems and ensuring the availability of quality data, for agentic AI to tap on. For AI agents to be useful, they need context and knowledge, which is best provided when organisations are strategically prepared to deploy this technology.

“To get knowledge, you need your systems and data to be integrated and available. And if organisations forget about these fundamentals, [scaling up to agentic AI] kind of loses its point,” he said.

Sometimes, AI agents will work best on the less visible sides of the organisation, such as helping to optimise operations, or on a smaller scale with simpler tasks.

Understanding the type and scale that works best according to the experiences and services that each organisation seeks to provide is key to taking advantage of the agents, said Ng.

Adopting a new mindset

Though agentic AI represents great promise for organisations, many public sector leaders believe that they are not ready to adopt it at the moment. They are primarily concerned with how to enable an agentic AI-ready workforce, and to lay the groundwork for future AI adoption.

According to Singapore’s Digital and Intelligence Service (DIS), Deputy Director for AI Labs, Ong Mei Xi, “workforce readiness strategy and regulatory certainty” are two key aspects that organisations must prepare to adopt agentic AI with confidence.

“Workforce readiness includes a strong part of culture, whether the workforce is culturally agile enough to adopt the mindset to be ready to try new things and new technologies,” Ong said.

In terms of regulatory certainty, she mentioned that there needs to be more robust guidelines on how to govern agentic AI at scale, in addition to existing frameworks like the AI Governance Framework.  

For instance, Workato’s GEARS framework emphasises on five domains representing organisational capabilities, including employee skills, processes, and technology operations. 

Speaking with GovInsider recently, Workato’s Ng highlighted the key understanding of value that organisations must be clear about agentic AI, in order not for it to turn into a “white elephant project”.  

Ng echoed this sentiment urging organisations to look inwards at the existing work processes where agentic AI has a role. “I like to think that where AI and agentic AI can really provide its best value or strongest value is on the boring stuff and on the invisible things,” he said. 

“It is also about the communication to the people that are going to work with AI agents, because we need to let them know that agents are here not to replace them, but in a way augment their work,” added JTC’s Wong. 

You can watch this full FOI panel recording here.