The White House wants quicker AI adoption. Can agencies make it happen?
The federal government is past the “should we use AI?” debate. The real question now is whether agencies can deploy it fast enough to deliver results.
Recent executive actions make that expectation clear. OMB Memorandum M-26-05, issued in January, calls for a more agile, risk-based approach to technology adoption. But while policy is accelerating, execution is not. On the ground, AI efforts are still slowed by fragmentation, redundancy, and long delivery cycles.
This is the core challenge: A widening gap between ambition and implementation. Closing it will require clearing structural bottlenecks in the IT approval process and confronting the cultural friction that keeps AI stuck in pilot mode.
Until agencies can scale beyond isolated use cases, the promise of AI-enabled government will remain more directive than reality.
The paradox of bureaucratic prudence
The execution gap is most visible in how agencies approve and deploy AI. Federal policy sets direction, but authority to act is spread across too many layers. This creates bottlenecks at the CIO and CDO level. These leaders are expected to deliver AI to mission teams, but they are often forced to evaluate tools that change weekly using approval processes built for slow, predictable software.
In many agencies, AI is still treated like a traditional IT product. That approach feels cautious, but it leads to long review cycles. Tools can spend months in evaluation, even as they continue to evolve. By the time they reach mission users, they are often already behind what is available in the commercial market.
One result is the rise of agency-built AI tools. To avoid procurement delays or security concerns, some IT teams create internal chatbots that resemble commercial products but offer far less capability. Many run on older models and are deployed well after newer versions are widely used outside government. These tools may meet compliance requirements, but they do not perform well for high-stakes mission work. They also frustrate employees who know better tools exist.
Recent OMB guidance on risk-based assurance is an important step forward. It gives CIOs and CDOs top cover to move faster without taking on unnecessary risk. In practice, this means they can tailor controls to the mission, approve limited deployments sooner, and update tools over time instead of locking requirements early. But flexibility alone will not close the gap. Agencies still need an execution strategy that supports AI continuously, integrates it into real workflows, and treats deployment as the starting point, not the finish line.
Closing organizational gaps and preparing the workforce
Beyond making tools available, AI adoption depends on whether people can use them in real work. In government settings, this is often difficult. Many employees are unfamiliar with approved tools. Others are unsure how to use them responsibly in daily workflows. At the same time, a small group of advanced users operates quietly. This hidden usage limits the agency’s ability to share proven, safe examples across teams.
Workforce hesitation is not only a training issue; it is also about trust and incentives. Some employees worry that using AI will make their roles seem less valuable. Others fear it could lead to job cuts. In an environment of constant budget scrutiny, efficiency can feel risky. Saving time does not always feel rewarded.
These challenges are compounded by how agencies are organized. Missions are not siloed, but authorities are. Mission leaders in areas like air traffic control or health care need help finding tools that fit their workflows and meet security requirements. Yet approval, security, and procurement functions are often separated from mission teams. When those connections break down, capable tools go unused or misunderstood. Closing this gap requires treating AI enablement as a shared service that supports mission delivery, not just an IT responsibility.
Why AI breaks the federal IT delivery model
Access to frontier models is no longer the primary hurdle; execution is. With 43 agencies already signed up for GSA USAi contracts, enterprise licenses are available for as little as $1. The Jan. 21 OneGov agreement with Broadcom further reduces costs while expanding access to AI-ready software.
To secure a true operational advantage, we must move from loosely distributed recommendations to clearer, centralized execution authority in shared mission areas like procurement and finance — while preserving flexibility for domain-specific missions.
We must also leverage new talent pipelines to close the gap between technology and mission execution. In addition to the OPM Data Science Fellows program launching this spring, the U.S. Tech Force initiative — a federal effort to recruit 1,000 technologists for time-limited placements focused on AI and modernization — signals a new approach to infusing agencies with technical experts and bridging workforce capacity with mission needs.
These professionals can help mission leaders navigate new tools and ensure they are integrated effectively into real workflows — not just approved in principle.
Execution is the modernization test
Bridging the gap between IT authorities and mission operators is now the defining challenge of federal AI adoption. Without execution, access becomes accumulation, and the government risks stockpiling tools instead of delivering results.
AI breaks the existing federal IT delivery model because it cannot be treated like stable enterprise software. It requires continuous enablement, workforce fluency, and mission-specific adaptation to produce real impact.
Meeting that reality demands new delivery approaches that embed AI into everyday workflows and prioritize speed alongside security. Modernization will ultimately be judged not by how widely AI is procured, but by how effectively it improves mission outcomes and public service.
Jesse Lambert is senior principal for strategy & AI adoption at Evans, a strategic management services firm.
