The rapid evolution of generative artificial intelligence has presented both remarkable opportunities and challenges in the public sector. Since ChatGPT burst onto the scene in 2022, followed by DeepSeek R1 in 2025, AI has evolved faster than many government agencies can adapt. Still, there’s no avoiding AI adoption in government – pressure is increasing for greater efficiency and improved services. Thus, public administrators are not wrestling with whether or not to adopt AI, but how to do it in the right way.
Success takes more than buying the latest software and bringing in technical experts. It demands a comprehensive approach to organizational change. Four key challenges stand in the way of successful AI implementation in the public sector: creating effective governance for these new technologies, building the right organizational capabilities, managing the inherent risks of these systems, and implementing AI in ways that support rather than replace human judgment.
A first relevant issue to be addressed is the introduction and effective use of AI tools in organizations. Infrastructure and technology do not automatically equate with improved performance because organizations are socio-technical systems – a set of relationships between humans, machines, and company culture. This sociotechnical perspective reveals three critical dimensions of AI implementation:
First, the human dimension. It is necessary to understand what capacities need to be developed and what resistances need to be overcome. When AI systems enter the workplace, they fundamentally alter employees’ roles and required skills (note, for good or for bad).
Second, the power dynamics. Technology changes the relationships between actors within organizations, as it eliminates positions and creates others, and transforms interactions and roles. For instance, the use of algorithms in some cases may limit the discretion of operators in their relations with users, shifting this power into the hands of those who define the criteria and instructions on which the algorithms are based. When an AI system rather than a caseworker determines benefit eligibility, decision authority moves from frontline staff to system designers.
Third, the cultural dimension. The introduction of AI interacts with the organizational culture, putting a strain on overall capabilities, openness to innovation, and project leadership. Public agencies with rigid hierarchies and risk-averse cultures often struggle most with AI adoption.
The citizen relationship adds another layer of complexity. The application of artificial intelligence in the production and delivery of public services implies building a level of trust so that users accept the role played by algorithms instead of humans. Without this trust, even technically excellent systems will face resistance.
The Joint Research Centre identifies factors driving successful AI adoption by public organizations: leadership support through incentives, an innovative culture open to new developments, a clear AI strategy with implementation guidance, and in-house AI experts spanning legal and ethical domains in addition to technical.
One of the main factors for the adoption, maintenance, and evolution of AI systems is the enrichment of in-house skills in public administrations. It is no longer possible to replicate methods of previous decades characterized by reliance on external providers alone. Today’s public sector needs internal capabilities to guide AI implementation, even if vendors provide the technical infrastructure. The introduction of digital technology requires managing complex internal socio-organizational processes and building relationships with networks outside the organization. It requires multi-faceted managerial skills.
Another research study conducted by the Joint Research Centre identified three essential skill categories for AI development in the public sector:
- Technical skills – Understanding AI capabilities, limitations, and implementation requirements.
- Managerial capabilities – Leading change, managing vendor relationships, and overseeing AI projects.
- Policy expertise – Related to legal aspects and ethics.
Focusing on the skills related to legal and ethical implications, there is now a clear awareness of the critical issues arising from AI systems. For instance, most chatbots are created by international companies that have designed proprietary algorithms without the models being known to the users (some companies keep the models and training methods confidential); furthermore, the way algorithms arrive at specific responses is often unknown. This creates significant challenges in determining who is responsible for the use of AI systems, privacy issues, profiling bans, and even the inappropriate use of systems by operators and cybersecurity management. Public administrators need expertise to be able to handle these complex issues.
Using AI systems comes with risks, making monitoring, auditing, and human oversight essential. While we understand the technology behind AI chatbots, how they generate responses remains unclear. These systems predict text probabilistically but do not actually “think,” which can lead to errors or nonsensical answers or “hallucinations.” This makes human supervision crucial.
AI also raises ethical concerns regarding fairness. AI decision-making has sometimes led to discrimination violations, as seen in the case of the use of algorithms in crime prevention in the United States. Transparency in algorithm design is necessary to prevent such issues.
Beyond ethical concerns, AI raises workforce questions. AI could impact jobs—replacing workers rather than enhancing productivity – potentially increasing inequality. Public sector leaders must consider how AI will affect their workforce and develop transition strategies.
Another risk is the loss of human discretion in decision-making. AI can help make decisions more consistent, reducing bias, but it must be carefully supervised to prevent mistakes. Automated decision-making (ADM) should be monitored to ensure fairness and accuracy.
Privacy concerns are also significant, especially when AI is used for profiling individuals, affecting personal rights and government relationships. These risks have prompted regulatory action, including the EU AI Act, and guidelines for AI use in the public sector.
Gone are the days when simply automating existing processes was enough.
AI adoption in the public sector must consider local needs and integration. While national governments play a key role in digital infrastructure (public platforms, cloud services, and cybersecurity), local administrations remain crucial due to their direct interaction with citizens. This dual-layer approach ensures AI systems address nationwide standards as well as community needs.
Smaller municipalities, often lacking resources, need support to keep up with larger cities in digital transformation. The digital divide isn’t just between citizens – it also exists between government entities themselves. When only well-resourced cities can implement advanced AI solutions, inequality in public service delivery increases.
A balanced approach to AI adoption should focus on territorial development, ensuring all regions benefit from new technologies. This enhances accessibility, well-being, and socio-economic cohesion. Strategic resource allocation and knowledge sharing between administrations can help here.
For AI to be effective in the public sector, it should follow a “pro-human” approach, complementing rather than replacing workers and citizens. This means designing systems that augment human capabilities. Gone are the days when simply automating existing processes was enough. The most successful AI implementations enhance both worker effectiveness and citizen experience.
Ultimately, successful AI implementation in government is not just a technical challenge – it’s a governance challenge that requires rethinking how public organizations operate at all levels, from national ministries to local service centers. Leaders who approach AI strategically, building the right capabilities and safeguards across all territorial levels, will create public services that are appropriately responsive, efficient, and equitable. We’ll all benefit from that.
© IE Insights.
