Why The Next Wave Of Agentic AI In Government Won’t Come From Silicon Valley

Written by Vitaliy Baklikov, CTO at Skyward IT Solutions.

Agentic AI is no longer a theoretical concept. In 2025, it’s becoming operational reality — especially within high-stakes sectors like healthcare, finance, and defence. As the Public Sector begins exploring agent-based systems to support mission-critical workflows, a key question is emerging: who is best positioned to build these agents?

While many associate innovation in AI with large, consumer-focused technology firms or startups, the needs of the public sector demand a different approach. In government operations, success depends less on scale and more on precision, compliance, and domain fluency. Agentic AI in these settings will likely be driven by specialised teams deeply embedded within the sector’s existing infrastructure.

Early AI Agents focused on broad applications: writing summaries, answering customer queries, or supporting developers. But in federal government environments, the opportunity lies in more complex, regulated workflows.

At Skyward, we’ve seen this first hand in our work with the Centres for Medicare & Medicaid Services (CMS). As one of the federal agencies, CMS operates 27 data warehouses and oversees 1 billion medical claims processed annually. Within such systems, the potential for Agentic AI is substantial — not only for clinical decision-making, but for automating operations, enhancing efficiency, and ultimately improving public service delivery.

Consider a federal agency unit responsible for generating weekly data reports. Previously, this task required a 10–12-person team, manually gathering data from federal websites, tracking legislative updates, and synthesising trends.

We built an AI Agent prototype capable of replicating that process with human-in-the-loop. The result: a job that used to take several days was reduced to minutes — without compromising quality or oversight. This is one small example of what’s possible when agents are designed with real-world workflows in mind. However, it’s important to note that this performance cannot be reached with an off-the-shelf agent. The narrower the scope of the agent, the faster a certain task can be executed. Open source models that are tailored to specific workflows are the fastest solutions for such tasks.

Agentic AI is now drawing serious attention, and market indicators reflect this momentum. According to industry estimates, the market grew from $5.1 billion in 2024 to $7.38 billion in 2025 — a 44.7% year-over-year increase. Projections suggest the sector could reach $47.1 billion by 2030. A considerable part of that growth will definitely happen in the public sector.

But this is not just about growth. It’s about readiness. Government agencies are actively exploring how agents can improve delivery without compromising compliance, security, or transparency.

The long-term vision isn’t just isolated agents; it’s interconnected systems that work together within a larger governance framework. Agencies are beginning to think in these terms, asking foundational questions:

• What governance structures should oversee AI agent deployment?
• How do agents running on different LLMs integrate securely?
• How can CIOs and CAIOs maintain visibility into agent usage and performance?

The answer may lie in creating modular ecosystems — federated models where agents can be safely developed, monitored, and improved across departments, with central oversight ensuring alignment with policy, ethics, and mission outcomes.

Public-sector AI must meet the highest standards. Agentic systems deployed in federal contexts must:

• Pass rigorous security reviews (ATO, FedRAMP)
• Be explainable by default, enabling human oversight and auditability
• Follow ethical design principles that prioritise safety, fairness, and accountability

These are not optional considerations — they are baseline requirements for adoption at scale

As Agentic AI continues to evolve, we anticipate three key shifts:

• Government teams will increasingly deploy domain-specific agents to automate niche but critical workflows
• Agencies will begin connecting these agents via federated systems of agents
• Chief AI Officers in partnership with Department Leads will emerge as pivotal figures in managing AI strategy, alongside CIOs and CISOs

The next chapter of AI adoption will not be driven solely by technological breakthroughs, but by thoughtful implementation in real-world settings. In government, this means focusing on domain specificity, compliance, and operational value.