Key Details
The World Bank’s How Is Government Using Artificial Intelligence? examines both the tools governments use and the conditions needed to use them effectively.
The survey covered 60 economies, drawing on responses from central digital agencies and officials in sectors including health, education, taxation and public finance.
44% of reported internal government AI use was ad hoc — individual staff using tools for tasks such as searching or summarising information.
Nearly two-thirds of surveyed governments provided public servants with official access or licences for generative AI. 39% had formal ministry-wide guidance on its use; another 38% were developing guidance.
India was not among the surveyed economies. The findings cannot be read as an assessment of Indian government AI use.
AI Has Entered Government Work, but Integration Is Limited
The paper finds AI in use across many surveyed governments, particularly for everyday staff assistance and citizen-facing chatbots. Tax administration shows more developed applications, including tools to flag potential fraud or answer taxpayer queries.
Access is different from integration. A public servant using AI to summarise a document is not the same as an agency incorporating it into a routine service with defined data, review and accountability arrangements. The study describes adoption as broad but often shallow on that measure.
Staff Guidance Has Not Kept Pace with Access
The gap between official tool access and formal guidance raises practical questions: what information may staff enter into a system, when must a person check its output, and who is accountable if AI informs a government decision? The paper identifies data quality, staff capability and organisational arrangements alongside guidance as conditions for effective use.
Barriers also vary. Surveyed lower-income governments more often cited missing policies and limited local-language capability; higher-income governments more often cited privacy and ethical concerns. A single model of public-sector AI readiness would therefore miss important differences.
Use Is Easier to Count Than Results
The survey records what governments say they have adopted, not whether those tools have improved services. Monitoring remains limited in many of the surveyed administrations and often tracks usage rather than service quality or costs. The paper’s contribution is a picture of how AI is entering government, not proof of gains in public-service performance.
Policy Relevance
For Indian ministries and states, the useful distinction is between giving staff AI tools and making an AI-assisted service dependable. The latter requires suitable data, clear staff guidance, human review where decisions affect people, and a way to assess whether the service has improved.
The study offers no India score or finding. Its relevance is as a test for Indian public agencies considering AI: can they identify the administrative problem being solved and measure the result, rather than treating the number of tools or users as the outcome?
Follow the Full Paper Here: How Is Government Using Artificial Intelligence?, World Bank Policy Research Working Paper 11466