How can data and AI drive greater public value?
B · Useful, limited EY and Oxford Economics survey of 492 senior government leaders in 14 countries with 46 interviews; method is disclosed, but results are self-assessed and the pioneer segmentation is EY's own index.
Summary based on the full text·Original: about 45 minutes to read
In short
The first of a two-part EY report on data and AI in government, based on a survey with Oxford Economics. Government leaders agree AI could cut costs and improve services, but few have deployed it. EY splits respondents into "pioneers" (top 20% on a readiness index) and "followers" and argues that pioneers built data infrastructure, digitised processes and analytics before moving to AI, and paid attention to workforce readiness at the same time.
Key numbers
- 26%government organisations that have deployed AI (excluding GenAI) partially or fully
- 12%organisations that have deployed GenAI partially or fully
- 64%respondents who agree AI adoption will bring significant cost savings and staff redeployment
- 2.4xhow much more likely pioneers are to report AI initiatives exceeding expectations
- 62%respondents citing privacy and security as a top constraint
As reported by the source.
English·PDF·opens ey.com
What it reports
- 64% expect significant cost savings and staff redeployment from AI; 63% say AI can improve services; 61% expect GenAI to change how government works within five years; 58% say adoption must speed up.
- Deployment lags belief: 26% have deployed AI and 12% GenAI, partially or fully; 4% have no plans to implement AI.
- Pioneers vs followers: data and digital infrastructure deployed 88% vs 58%; processes digitised 58% vs 33%; analytics deployed 76% vs 33%. The gap narrows for AI and GenAI themselves.
- Pioneers are 2.4x more likely to report AI initiatives exceeding expectations.
- Top constraints: privacy and security (62%), strategic misalignment (51%), inadequate infrastructure (45%), ethics (42%), weak business cases (41%).
How they got it
Survey of 492 government leaders (CIOs, CDOs, chief AI officers and similar) in 14 countries across national, regional, local and state-owned bodies, plus 46 interviews (38 public officials). Pioneers were identified with an equally weighted index on planning, metrics, workforce measures and infrastructure, normalised 0 to 100. EY used its own LLM tools and Microsoft Copilot to synthesise interview transcripts.
Limits
- All adoption and outcome figures are self-assessed; EY names this as a source of bias.
- The pioneer index rewards foundations, so the finding that pioneers prioritise foundations is partly built into the definition.
- "Exceeding expectations" is a perception measure, with no cost or service data.
- Published as a lead-in to EY advisory services; the practical guidance is deferred to a second report.
Why it matters for companies
The sequencing finding transfers beyond the public sector: organisations that report AI success did the data, process digitisation and analytics work first. For anyone building an AI case inside a large organisation, the barrier list is a usable checklist, with privacy and unclear business cases ahead of technology. Treat the pioneer premium as directional, since it comes from the same respondents who rated their own success.
Summary based on the full text. Read the original for the full argument and data.Open PDF on ey.com