AI in the government sector 2026
C · Weak signal A KPMG Singapore sales deck that compiles press reports, vendor figures and two earlier KPMG surveys, with no original analysis and two case studies written to promote its own services.
Summary based on the full text·Original: about 20 minutes to read
In short
A 20-page deck from KPMG Services in Singapore (May 2026) on AI in government and public education, centred on Singapore's Smart Nation 2.0 and National AI Strategy 2.0. It lists trends, opportunities and risks, gathers figures from press and agency sources, summarises two KPMG global surveys of public-sector leaders, and closes with KPMG case studies and service offerings. The main message: optimism about AI in government is high, but operating models, legacy systems, budgets and cyber risk are slowing scale.
Key numbers
- ~30,000cases per month handled by Singapore GovTech's OneService chatbot, which predicts case category and routes to agencies
- up to 50%call deflection attributed to the Ask Jamie assistant across government sites, after more than 15.0 mn citizen questions
- 16%government respondents who say their operating model consistently enables AI agents in workflows
- 75%Singapore teachers using AI, against 16% of lower secondary teachers using it to support students with special educational needs
As reported by the source.
English·PDF·opens assets.kpmg.com
What it reports
- Singapore planned about S$3.3bn of government ICT spending in FY2023, with more than 30.0% (about S$1.0bn) for applications on the Government Commercial Cloud.
- Singapore committed about $743.0 million over five years to AI under NAIS 2.0.
- GovTech's OneService chatbot has run since 2021 and handles about 30,000 cases a month; Ask Jamie has answered more than 15.0 mn questions with up to 50% call deflection.
- From KPMG surveys (115 and 184 public-sector respondents): 84% expect AI to shift from efficiency to strategic innovation within a year; 79% believe AI will give governments a significant advantage; 63% have used AI for under two years; only 16% say their operating model consistently enables agents in workflows, and 12% say AI is integral to their organisation.
- Education: 75% of Singapore teachers use AI; 82% say it helps lesson planning and 74% that it automates admin; teachers still work about 47 hours a week. NTU makes AI literacy mandatory for undergraduates from August 2026.
- Risks cited: 201 public-sector data incidents in 2023 (up 10%); the 2018 SingHealth breach exposed 1.5 million patient records.
- A KPMG penetration test of a government LLM application showed prompt injection could make it delete or alter read-only user databases and bypass content restrictions.
- Case study: Covered California's bilingual virtual assistant (built for 2019 open enrollment, 90+ common queries) reduced peak call-centre load; no figures are given.
How they got it
Desk research from secondary sources (BCG, GovTech, Straits Times, market-research firms, Microsoft press releases) accessed in March and April 2026, plus key findings lifted from two KPMG survey reports. There is no stated method for selecting trends or examples.
Limits
- Several figures are old (FY2023 spending, 2018 breach) or vendor-reported (chatbot deflection, Microsoft investment).
- Case studies report no measured outcomes and serve as service promotion; the California assistant dates from 2019.
- Focus is almost entirely Singapore, so transfer to other systems is not shown.
Why it matters for companies
The deck shows how a mature digital government frames AI: shared platforms, a common cloud, and governance tooling built before scaling. Two items are directly usable in a company. The penetration-test case is a clear example of "excessive agency" risk: an LLM tool with write access can be talked into destroying data. And the teacher data (high adoption, unchanged 47-hour weeks) is a reminder that AI use and workload reduction are separate outcomes that need separate measurement.
Summary based on the full text. Read the original for the full argument and data.Open PDF on assets.kpmg.com