Opportunities and risks in public sector use of GenAI
By Amit Roy Choudhury
Speakers from IMDA, Changi General Hospital, Synapxe and IBM shared about both the immense opportunity GenAI presents in improving public service as well as its potential risks during GovInsider’s FOI.
Presenters from IMDA, Changi General Hospital, Synapxe and IBM shared about the possibilities as well as challenges in the use of GenAI by the public sector to provide better service. Image: GovInsider.
Earlier this year, Singapore’s Minister for Digital Development and Information, Josephine Teo, at an event in France highlighted Singapore’s commitment to using artificial intelligence (AI) to improve lives.
Continuing on this theme of the public sector using AI and generative AI (GenAI) for good, speakers at a session titled GenAI in Government Services: Opportunities, Risks, and Strategies during GovInsider’s recent Festival of Innovation, shared the approach being taken by their organisations in the use of GenAI.
Here is a compilation of some of the major points made by these speakers.
Janet Chiew, Deputy Director, Incubation, Infocomm Media Development Authority (IMDA), Singapore
Chiew, who is from IMDA’s Business Tech Group, shared about GPT Legal, a GenAI model fine-tuned for the legal sector.
She said GPT Legal sought to address the unique challenges of Singapore's legal landscape by leveraging GenAI’s potential in the legal sector.
Chiew said the model has been designed to understand and process legal terminology, documents, as well as case laws unique to Singapore's complex legal system - which incorporates elements from English common law, US Intellectual Property (IP) law, and the Indian Penal Code.
The GPT’s initial focus was on legal document summarisation, with the ability to reduce the time required to summarise judgments from two days to approximately 10 minutes.
IMDA has implemented sophisticated safeguards to address potential AI hallucinations, including accuracy checks at entity and paragraph levels, as well as a unique "fact score" metric that helps users understand the reliability of generated summaries.
Future plans for GPT Legal focused on expanding the model's capabilities beyond summarisation and deepening its understanding of Singapore's unique legal ecosystem, she added.
The objective was to collaborate with industry partners and legal tech firms to develop transformative tools that could uplift job scopes and roles across different legal practice areas.
This included potential applications in front-office technology, back-end services, and comprehensive support for lawyers in various legal domains, she said.
The full video of Janet Chiew’s presentation can be found here.
Dr Charlene J Y Liew, Deputy Chief Medical Informatics Officer, Changi General Hospital, Singapore
Dr Liew highlighted four key barriers to realising the full potential of digital healthcare.
One of them was that healthcare systems currently lack the comprehensive AI infrastructure needed to fully integrate digital technologies.
Likewise, legacy IT systems that were decades old created significant barriers to seamless digital healthcare implementation.
To subscribe to the GovInsider bulletin, click here.
Dr Liew added that current care pathways were designed in an "AI-naive" era and needed to be redesigned to be "AI-indicative", considering new technological capabilities.
The final challenge was that there was a critical need to upskill healthcare professionals to create an AI-literate workforce.
Dr Liew said simply digitalising existing inefficient workflows did not work as it just created “digitally inefficient workflows”.
According to her, the use of Agentic AI represented a transformative leap in healthcare technology, offering the potential to autonomously complete complex tasks and orchestrate information from multiple sources.
“By breaking down information silos, Agentic AI can synthesise data rapidly, presenting clinicians with comprehensive, contextualised insights that could dramatically enhance diagnostic accuracy and treatment planning,” she said.
Dr Liew, however, added that the promise of Agentic AI in healthcare was tempered by critical considerations of safety, reliability, and human oversight.
She stressed the point that AI will not replace doctors, but doctors who use AI will replace those who do not.
The full video of Dr Charlene J Y Liew’s presentation can be found here.
Dr Goh Han Leong, Senior Principal Specialist, Data Analytics & AI, Synapxe, Singapore
Dr Goh said Synapxe, as Singapore's national HealthTech agency, was pioneering the use of GenAI and advanced technologies to transform healthcare delivery.
In this context, he mentioned projects like Health Kaki, which is a GenAI-powered health companion designed to empower residents with personalised health plans that align with their individual goals and preferences.
He noted that by leveraging large language models (LLMs) and an Agentic AI framework, Synapxe was looking to move beyond a traditional one-size-fits-all approach to healthcare.
During his presentation, Dr Goh emphasised the difference in scope between AI used in population health and clinical AI.
Clinical AI focuses on improving the health outcomes of individual patients, targeting specific medical interventions and personalised treatment strategies, while the goal of population health AI is to implement strategies that can improve health outcomes at a national level, he said.
Dr Goh gave the example of the Healthier SG scheme as a critical case study of the use of population health AI.
The scheme emphasised preventive care, patient empowerment, and creating scalable health interventions that can be applied across diverse population segments.
The full video of Dr Goh Han Leong’s presentation is available here.
Su Yin Anand, Lead Client Partner, Healthcare Consulting, IBM
Taking a different tack from the previous speakers, Anand noted that according to Gartner research, 85 per cent of AI pilots fail to mature, and this was a major concern.
She noted that AI pilots often fail due to misalignments between technological implementation and strategic business objectives.
“Many organisations launch pilots without a clear understanding of how AI should integrate into their existing workflows, treating it as a standalone solution rather than a transformative tool,” she said.
Anand noted that AI governance was the critical foundation for building trustworthy and scalable AI systems.
“Effective governance goes beyond traditional compliance, requiring an automated approach to model lifecycle management that can keep pace with rapidly evolving AI technologies,” she said.
Anand added that organisations needed to establish an AI ethics board, defining clear guiding principles for model selection and use, and implementing robust mechanisms to monitor AI models' continuous learning processes.
The goal is to transform AI from a potentially risky technology into a reliable, accountable tool that can be confidently integrated across different organisational functions, Anand said.
The full video of Su Yin Anand’s presentation can be found here.