by Paul Rempfer
Government process modernization is falling behind the pace of AI adoption, and that gap is where programs start to break. When agencies automate workflows that were originally built for paper files, manual review, and siloed authority, there’s little to no modernization taking place. In fact these efforts are really just scaling inconsistency. The Government Accountability Office (GAO) has been warning about this pattern across major programs: new technology layered onto old ways of working produces rework, delays, and outcomes that are harder to explain or defend.
I’ve spent more than three decades working in cyber, intelligence, and national security, leading missions and modernization efforts across federal agencies and allied governments. Across every major technology shift, I’ve seen the same pattern. When agencies layer new tools onto legacy processes, they become more complex, harder to manage, and harder to defend.
This piece builds on that reality. In the sections ahead, I’ll lay out why AI is amplifying legacy process debt across agencies, show what oversight bodies have already flagged in real programs, and offer a practical, process team approach for redesigning mission workflows.
The Core Problem: AI Is Being Added to Systems Built for a Different Era
Most federal processes were built for a world of paper files, manual review, and human judgment at every step. Over decades, they accreted new laws, oversight requirements, and reporting obligations. What they rarely became, however, were coherent end-to-end systems with shared data, consistent decision logic, or clear ownership of outcomes.
When AI is layered onto that foundation without redesigning the workflow underneath, it does not modernize the process. It accelerates individual steps while leaving structural fragmentation intact. The result is faster throughput inside systems that are still inconsistent, duplicative, and hard to govern.
This pattern is visible across the government.
Fragmented workflows produce inconsistent outcomes
At the Internal Revenue Service (IRS), modernization efforts have expanded automation and analytics while core case-handling workflows remain fragmented across systems and offices. During the 2023 filing season, the IRS answered fewer than one-third of inbound calls at peak times, even as new AI tools were introduced to improve service. GAO and Treasury Inspector General reviews have shown that taxpayers often receive different answers depending on which channel or office they reach, because the underlying workflows and authorities are not standardized. AI in that environment inherits inconsistency rather than eliminating it.
A similar dynamic appears at the Social Security Administration (SSA). The agency introduced AI-assisted document classification and case-support tools to help address disability and benefits backlogs. Yet average initial disability decisions still frequently exceed 200 days. Automation surfaces issues faster, but frontline staff often lack the system access or decision authority needed to resolve them. The bottleneck is the process wrapped around the model vs. the model itself.
Standardization failures turn modernization into operational risk
The Department of Veterans Affairs (VA) provides one of the clearest warnings. Before deploying its electronic health record (EHR) modernization program, the VA did not standardize clinical workflows or documentation practices across facilities. Each medical center operated with heavily customized processes. When those nonstandard workflows were forced into a single modern system, the result was widespread disruption rather than improvement.
Inspector General and GAO reviews documented serious operational failures, including patient safety risks. As late as 2024, 75% of users reported that the new EHR made them less efficient, not more. The technology was not the core failure. The failure was modernizing systems without first modernizing the work.
Automation without authority creates parallel decision paths
Across benefits agencies and contact centers, AI is increasingly used to classify documents, route cases, and flag issues. But when escalation authority, data access, and accountability remain fragmented, automation creates parallel paths rather than a single clear decision flow.
Contact center staff may see problems faster, but cannot resolve them. AI-generated insights move cases forward, but human reviewers still operate under legacy approval chains. The organization ends up doing the same work twice, once by software and once by people, with no clear owner for the final outcome.
Defense missions stall on process
Inside the Department of Defense, senior leaders have been explicit about the constraint. AI is not being limited by model availability or computing power. It is being limited by governance, data ownership, and inconsistent processes.
Logistics, personnel management, and mission-support environments remain constrained by Impact Level requirements and fragmented workflows. Service desks and operational support functions still rely on handoffs designed for a pre-AI era. Until those processes are standardized and redesigned, scaling AI safely and effectively remains out of reach.
The Organizational Risk Most Agencies Are Underestimating
The environment I’m describing is not just an IT issue. The real challenge is both workforce and organizational in nature, one that federal oversight bodies have been warning about for years.
Large portions of the federal workforce exist to review information, reconcile data, and move cases across organizational seams. Those are precisely the tasks AI performs well. Agencies know this, but they are not fully implementing workforce planning activities needed to manage this transition.
When AI is introduced without redesigning authority, roles, and accountability, it creates a quiet but dangerous situation:
- People stay in roles that slowly matter less.
- Offices continue to exist even though the work they were created to do is now handled by software.
- Employees see AI as a threat rather than a support tool because no one is explaining what their future role will be.
Over time, this erodes morale, increases resistance, and weakens mission performance.
7 Practical Steps to Redesigning Mission Workflows
AI does not improve operations on its own. It only accelerates whatever work structure already exists. If agencies want AI to support agile delivery, IT modernization, and workforce planning, they have to redesign how work actually flows before they automate any part of it.
1. Stand up a dedicated process function.
Create a small, permanent group of business process specialists who partner with mission owners to make the work itself clear, consistent, and ready for automation. I’m not suggesting creating an innovation lab or a data science shop. It should be a core management function, just like finance or human resources.
2. Map reality, not policy.
The process teams will sit with the mission teams and follow cases end-to-end. They’ll observe and document who does what, where decisions happen, where work stalls, and which systems and handoffs drive rework. They’ll ask hard questions about relevance. Which steps exist only because they always have? Which offices exist mainly to move information from one place to another? Which roles are likely to shrink once automation is in place?
This information allows them to build one shared “current state” of how the organization actually operates and make comparisons across an agency. For example, when two offices are doing the same kind of work in different ways, leadership should be able to see that clearly. When three organizations exist mainly to support the same step in a workflow, that should be visible rather than hidden in job descriptions.
3. Simplify before you automate.
The process team’s next job may be where the most modernization value is unlocked. They’ll simplify their findings, strip out duplicate reviews, value-less handoffs, and steps that exist only to compensate for broken systems. They would separate what is legally required from what is tradition.
4. Redesign for AI-enabled execution.
Only after simplification can process teams redesign a new version of the workflow for today. In that design, routine rules would be automated and AI would be used for sorting, prioritizing, and pattern recognition. Humans would still be in the loop to focus on exceptions and provide high-impact judgment and accountability. Process teams would collapse parallel decision paths into one clear route with defined escalation points.
5. Align systems, data, and accountability.
Government IT modernization then configures to match the redesigned workflow instead of the old one. This is where technology purchasing decisions are made, and process teams can reduce copying and reconciliations by clarifying data ownership and sharing. They’d also ensure automated decisions are logged and traceable so an agency can explain outcomes under review.
6. Manage workforce change on purpose.
Given their full picture understanding, process teams can help identify which roles change, which shrink, and which become oversight functions. They can work with HR and agency leadership to define new roles focused on oversight and judgement and make workforce transition a planned activity, not an aftershock.
7. Sustain the workflow so modernization sticks.
This last step is where government process modernization becomes durable, not a one-time redesign. The process team stays engaged to track cycle time, rework, and where humans have to intervene. They monitor where exceptions spike, where handoffs creep back in, and where new AI tools are quietly creating parallel paths. Then they tune the workflow, update the decision logic, and keep accountability clear as missions, policy, and systems change.
Agencies that follow these steps and engage process teams stop building parallel systems and overlapping offices. They reduce duplication instead of digitizing it. They move faster because work is designed for automation instead of being forced into it. Employees will know how their roles are expected to evolve instead of guessing. Resistance will be lower because change will be visible and planned. Most importantly, agencies stay connected to mission value: offices don’t linger after their purpose fades, and AI supports human judgment instead of quietly replacing it without a plan.
