Why Generative AI Hasn’t Yet Transformed Government — And What Needs to Change

Amid the growing excitement surrounding generative AI, it’s easy to assume that a technological revolution is already sweeping through government institutions. But beneath the surface of glossy headlines and policy reports, the real picture is far more restrained.

In a powerful essay titled “Why Generative AI Isn’t Transforming Government (Yet)”, public-sector innovation expert Tiago C. Peixoto offers a reality check. His diagnosis is not of failed potential, but of structural inertia, conceptual confusion, and governance misalignment. Generative AI may be promising, he argues—but the state is not yet ready to receive it.

From Capability to Impact: The Missing Link

Peixoto’s central thesis is clear: generative AI is not a plug-and-play solution. The public sector’s enthusiasm has outpaced its institutional capacity. Many so-called “AI deployments” in government turn out to be traditional automation or rules-based systems, not true GenAI implementations. The truly generative use cases—such as Brazil’s MARIA and France’s Albert—remain assistive, not transformative. They draft, summarize, and support. They do not decide, automate, or structurally change workflows.

This cautious approach is no accident. Governments, rightly, are built on predictability, accountability, and rule-based processes. GenAI, on the other hand, thrives on improvisation, pattern recognition, and statistical inference. This mismatch—between a jazz-like system and a bureaucratic orchestra—makes deep integration risky.

The “RAG Trap” and the Illusion of Automation

Peixoto points to a widespread reliance on retrieval-augmented generation (RAG)—a technique where large language models are combined with institutional databases to answer questions more reliably. These tools, while helpful, simulate transformation without delivering it. They don’t automate decisions, trigger backend processes, or update official records. They inform but don’t act.

This creates an illusion of modernization. Internally, systems look smarter; externally, the user experience remains unchanged. Without integration into operational systems and accountability structures, GenAI remains a detached assistant—not a civil servant.

Beyond Futurism: The Case for “Strategic Augmentation”

One of the most compelling contributions of Peixoto’s essay is his reframing of the value proposition: not automation for its own sake, but strategic augmentation. In bureaucracies plagued by caseload overload, cognitive bottlenecks, or inconsistent service delivery, GenAI can act as a force multiplier—supporting frontline staff, extending reach, and enabling faster response times.

In settings where qualified human resources are scarce—such as health clinics in sub-Saharan Africa or underfunded education systems—delegated autonomy may even prove superior to the status quo. The challenge is to design systems where GenAI operates within tight boundaries, under human oversight, and with clear fallback protocols.

Practical Steps Toward Readiness

Peixoto offers a roadmap to move governments from rhetoric to readiness:

• Disaggregate AI types and match them to high-value tasks, avoiding a one-size-fits-all narrative.

• Benchmark GenAI against real-world human performance, not idealized standards.

• Design pilots within actual workflows, and evaluate them against measurable service outcomes.

• Invest in adaptable governance, including participatory oversight mechanisms like civic red-teaming and dynamic audit frameworks.

• Modernize procurement and budgeting models, shifting from static capital expenses to iterative, service-based investments.

• Foster co-creation with model providers, customizing GenAI for domain-specific constraints, ethics, and performance needs.

Toward a Responsive, Not Just Efficient, State

Ultimately, the essay doesn’t reject the role of GenAI in government—it refines it. The goal is not automation for efficiency alone, but augmentation for equity, inclusion, and responsiveness. In this vision, GenAI serves not as a replacement for public servants but as a lever to extend their impact, especially in areas where the state is weakest or most stretched.

The real challenge isn’t technological; it’s institutional. Governments must evolve their accountability frameworks, legal architectures, and implementation cultures to responsibly harness this general-purpose technology. That requires political will, regulatory foresight, and most of all, humility.

As Peixoto concludes, GenAI’s true promise lies not in replacing the state—but in helping it serve better.

Source: Tec Policy