Public sector leaders globally are moving from exploring AI capabilities toward implementing practical solutions, with agentic AI increasingly being considered as a key approach. According to a global study commissioned by Dell Technologies and conducted by International Data Corporation (IDC), 71% of government decision-makers identify agentic AI—defined as systems that can autonomously carry out complex tasks—as an important factor in advancing AI adoption in government operations.
The findings point to several ongoing trends in public sector AI adoption:
- Agentic AI as a workforce capability tool: 51% of surveyed leaders report plans to invest in agentic AI within the next 12 to 18 months. This indicates a shift toward deploying autonomous systems for complex administrative and operational tasks, with the aim of reallocating staff time toward other responsibilities.
- Skills and capability gaps: Around 66% of public sector organisations report a skills gap, noting that technological change is progressing faster than workforce readiness. This is contributing to operational pressures and increasing interest in AI-supported systems.
- Requirements for safeguards and controls: While interest in agentic AI is increasing, 44% of leaders indicate they would only accelerate adoption if appropriate measures are in place around data security, privacy, and data sovereignty.
- Role of partnerships: 61% of respondents agree that public-private partnerships are important for accessing the expertise and technology needed for secure AI implementation.
The study also reports that 58% of leaders view strong data governance, data quality, and control as essential for effective AI deployment. Related concerns include ensuring data sovereignty and establishing governance frameworks that support scaling AI systems.
Overall, the findings describe a movement toward broader consideration and deployment of agentic AI in government settings, alongside continued emphasis on governance, security, and workforce capability as key factors influencing implementation.
