Written Answer · 2026-08-04 · Parliament 15
Measuring Productivity Gains, Service Improvements and Risks from Generative AI Deployments across Public Service
PAP MP Saktiandi Supaat asked the Minister for Digital Development and Information in a written question what the principal AI use cases across the Public Service are, how many public agencies have deployed generative AI systems in their operations, and how the Government measures the productivity gains, service improvements and risks arising from such deployments. Minister Josephine Teo, answering together with his oral question that day and a written question for the next sitting, said the Public Service is progressively deploying AI to solve operational problems and improve service delivery. Almost 70% of public officers use AI tools regularly to analyse information, prepare drafts and support document processing, and agencies are using AI to improve services for citizens. Adoption is broadening but its depth and maturity remain uneven across agencies; the current focus is learning from practical deployment and experimentation — where AI delivers the greatest value, how it affects work and teams, and what safeguards are needed. Because AI can be deployed through common platforms, embedded in larger digital services or adapted for agency-specific workflows, a simple count of AI systems would not be a meaningful measure of progress. Instead, agencies assess the benefits, costs and risks of deployments, with success judged by whether AI helps officers work better and deliver more effective and efficient services, within Government-wide requirements for procurement, governance and security. Agencies must implement safeguards and human oversight, and public officers are being trained to use AI confidently, effectively and responsibly.
Why it matters
Almost 70% of public officers already use AI tools regularly, yet the Government refuses to measure progress by counting systems, leaving agencies to self-assess benefits, costs and risks while adoption depth remains uneven.
Key Points
- • Almost 70% of public officers use AI tools regularly to analyse information, prepare drafts and support document processing, and agencies use AI to improve citizen services
- • Adoption is broadening, but depth and maturity remain uneven across agencies; the current focus is learning from practical deployment and experimentation
- • The Government says a simple count of AI systems would not be a meaningful measure; agencies assess the benefits, costs and risks of their own deployments
- • Success is measured by whether AI helps officers work better and deliver more effective, efficient services, within Government-wide procurement, governance and security requirements
- • Agencies must implement appropriate safeguards and human oversight, while public officers are trained to use AI confidently, effectively and responsibly
The Government positions Public Service AI adoption as "broad rollout, learning by doing": it cites almost 70% of public officers using AI tools regularly as evidence of reach while conceding that depth and maturity vary across agencies. It explicitly declines to answer the MP's request for quantitative measures with a count of AI systems or of agencies using generative AI, relying instead on agencies' own assessments of benefits, costs and risks and on ultimate service outcomes, with safeguards, human oversight and officer training running in parallel.
The questioner, Saktiandi Supaat, sought verifiable figures from the Government: the principal AI use cases across the Public Service, how many agencies have deployed generative AI, and how productivity gains, service improvements and risks are quantified. His concern was whether public-sector AI investment has measurable returns and risk controls; the reply did not provide an agency count or specific metrics.
The Government has formally stated that it will not use the number of AI systems as the yardstick for Public Service AI progress, devolving assessment to individual agencies with service outcomes as the final measure. This signals that Singapore's public-sector AI governance will not, in the near term, adopt a unified quantitative KPI or publish a deployment inventory; the emphasis is on experimentation within Government-wide procurement, governance and security requirements, backed by the trio of safeguards, human oversight and officer training.
"Given the many ways AI can be deployed, whether through common platforms, embedded within larger digital services or adapted for agency-specific workflows, a simple count of AI systems would not provide a meaningful measure of progress."
Participants (2)
Original Text (English)
SPRS Hansard · Fetched: 2026-09-04
75 Mr Saktiandi Supaat asked the Minister for Digital Development and Information (a) what are the principal use cases in which artificial intelligence is currently deployed across the Public Service; (b) how many public agencies have deployed generative AI systems in their operations; and (c) how the Government measures the productivity gains, service improvements and risks arising from such deployments.
Mrs Josephine Teo : My reply will address the Oral question raised by Mr Saktiandi Supaat in today's Order Paper, as well as his Written Question for the next Sitting, as they both relate to artificial intelligence (AI) in the Public Service.
The Public Service is progressively deploying AI to solve operational problems and improve public service delivery across a wide range of areas. Almost 70% of public officers use AI tools on a regular basis to analyse information, prepare drafts and support document processing. Agencies are also using AI to improve service delivery for citizens.
While adoption is broadening, the depth and maturity of AI adoption remain uneven across agencies. The current focus is on learning from practical deployment and experimentation, including where AI delivers the greatest value, how they affect work and teams, and what safeguards are needed to ensure responsible use.
Given the many ways AI can be deployed, whether through common platforms, embedded within larger digital services or adapted for agency-specific workflows, a simple count of AI systems would not provide a meaningful measure of progress. Instead, agencies assess the benefits, costs and risk of AI deployments, with success ultimately measured by whether AI helps officers work better and deliver more effective and efficient services to our citizens and businesses. As agencies continue to assess and refine their AI use cases, they do so within Government-wide requirements for procurement, governance and security.
Agencies must implement appropriate safeguards and human oversight for AI deployments and apply AI prudently to maximise benefits. At the same time, we are also training and upskilling public officers to use AI confidently, effectively and responsibly.