Reimagining the Federal Workforce - Dynamic PDs, Team Portfolios, and the AI Dividend

“The tools got smarter. The job evolved. The PD? That stayed in the drawer.” 

In Part 1 of this mini-series, Beyond the Chatbot, we confronted the gap between “Magic AI” expectations and public-sector reality. Real AI integration is not about dropping a commercial chatbot onto an intranet portal. It requires a value pipeline: data, context, security, embedded workflows, and human adaptation before AI produces mission value. 

But when those pieces come together, something more consequential happens. 

The job changes. 

Consider the program analyst from Part 1, reviewing grant applications with the help of an AI-enabled workflow. Her tools no longer require her to produce a first draft from scratch. Instead, she evaluates AI-generated work, checks it against law and policy, identifies gaps, frames the next question, and exercises judgment about what the machine cannot know. 

She is still a program analyst, but much of the work that creates value is different. 

That creates an institutional problem. We are introducing intelligence-age tools into federal agencies while much of our human-capital infrastructure remains designed around an industrial-age conception of work: define a position, enumerate its duties, assign a classification, and evaluate the employee against those responsibilities. 

If AI changes the work, shouldn’t we change the way we design, organize, and reward that work? 

The “PD in a Drawer” Problem 

The federal classification system established by the Classification Act of 1949 was designed for a very different environment—one in which work could be more readily decomposed into defined duties and associated with relatively stable positions and classifications. 

That model still performs important functions. But it becomes increasingly strained when the work itself changes faster than the documents that describe it. Traditional Position Descriptions can mask actual duties. They may remain unchanged while workflows, technologies, and mission demands evolve. 

They can also create an efficiency penalty. If an employee uses modern tools to complete an eight-hour routine task in two hours, what happens to the six hours saved? If the answer is simply “more of the same work,” employees have little incentive to invest in learning or adopting better tools. Efficiency becomes a way to increase workload rather than create capacity. 

And traditional PDs can obscure the fact that modern work is increasingly team-based and complementary. A position description describes an individual position. Mission performance depends on how those positions work together. 

The question is not whether we should eliminate PDs. It is whether we should supplement the static description of a position with a more dynamic picture of the capabilities, capacity, tools, and relationships that produce mission value. 

From Positions to Portfolios 

Decades before generative AI, organizational psychologist Meredith Belbin argued that high-performing teams depend on complementary human behaviors. His nine team roles recognized something important: effective teams do not require every individual to possess every capability. They require the team, collectively, to have the capabilities needed to succeed. 

AI introduces a new dimension to that idea. The question is no longer simply: Which human capabilities should we combine? It is also: Which capabilities should be performed by humans, which should be AI-assisted, and where should humans and AI work together? 

Return to our grants analyst. In the traditional model, she might build a model, a statistician might interpret it, and a policy expert might write the guidance—three specialized functions that can become three silos. 

In an AI-enabled workflow, routine coding and baseline data compilation may require far less human effort. That changes where human value lies: validating outputs, checking legal and ethical constraints, curating context, recognizing anomalies, managing stakeholders, and deciding what questions are worth asking. 

AI does not eliminate the need for complementary talent. It changes the portfolio of capabilities the team needs. 

That has implications for supervisors as well. Their role increasingly becomes less about distributing repetitive tasks to individual job codes and more about orchestrating a team of complements—matching human strengths such as empathy, negotiation, institutional knowledge, and domain intuition with increasingly capable AI tools. Call this the Talent Portfolio Manager role. 

The Dynamic Position Profile 

To support that model, agencies should consider supplementing static Position Descriptions with Dynamic Position Profiles built around four elements. 

1. Capability and Capacity Mapping 

A Dynamic Profile would capture the employee’s current mix of capabilities—their “skill vector”—rather than relying primarily on degrees, tenure, or years in grade. That might include stakeholder facilitation, policy synthesis, data analysis, tool fluency, prompt engineering, or other capabilities relevant to the mission. It could also provide visibility into current workload and available capacity. The objective is to give managers a better understanding of what the team can actually do now. 

2. The Team Portfolio Matrix 

Work rarely happens in isolation. A Dynamic Profile should therefore connect individual capabilities to team capabilities. Which skills are abundant? Which are scarce? Where are there single points of failure? For example, what if only one analyst understands the compliance guardrails embedded in a legacy system? A team portfolio makes those dependencies visible and allows leaders to rebalance work, develop bench strength, and reduce organizational fragility. 

3. Tool Fluency and Executive Editing 

In an AI-enabled environment, using the tool is only part of the job. Knowing when to trust it, when to challenge it, and how to improve its output becomes a professional competency. 

Core responsibilities may increasingly include: 

• Executive editing and auditing: evaluating, fact-checking, and calibrating AI-assisted work against legal, policy, and ethical requirements. 

• Tool literacy and adaptation: maintaining currency with evolving AI tools and workflows. 

• Problem framing: defining the context and questions that determine whether AI produces useful results. 

The emerging employee is not simply a faster drafter. The employee becomes the editor, validator, contextualizer, and decision partner. 

4. The AI Dividend 

This may be the most important element. If technology reduces the time required to perform routine work, the resulting capacity should not automatically disappear into a larger volume of routine work. Organizations should deliberately capture a portion of that productivity gain as an AI Dividend, a protected capacity reinvested in higher-value human activity. 

A reasonable starting management target might be to protect 10–15 percent of working hours recovered through automation for strategic process improvement, upskilling, cross-functional problem solving, experimentation, or career development. 

The precise percentage matters less than the principle: Productivity gains should create capacity, not merely raise the workload ceiling. Without such a mechanism, agencies may discover an unfortunate paradox: the more effectively employees use AI, the more work they receive. That is not an adoption strategy. It is an efficiency penalty. 

Rethinking Performance 

It is difficult to ask employees to innovate with AI when performance systems continue to emphasize raw production volume over cognitive quality. 

Suppose an analyst uses AI to synthesize 500 public comments in an hour rather than twenty. Measuring success primarily by pages produced per day sends the wrong signal. The employee has either an incentive to hide the efficiency gain or an expectation that the reward for working smarter will simply be more work. 

In the intelligence age, performance assessment should place greater weight on higher-order contributions: 

• Auditing rigor: Did the employee identify hallucinations, policy conflicts, errors, or bias? 

• Problem framing: Did the employee ask the contextual questions that led to accurate and actionable results? 

• Workflow optimization: Did the employee develop or share prompts, processes, or configurations that improved the team’s performance? 

• Knowledge transfer: Did the employee turn an individual productivity improvement into an organizational capability? 

The goal is not to reward employees for using AI. It is to reward them for creating more mission value because they know how to use AI well. 

Monday Morning: What Can Leaders Do Now? 

None of this requires waiting for a sweeping transformation of the federal classification system. Leaders can begin within existing flexibilities. 

• Map the team. Identify individual strengths, critical knowledge, tool fluency, and vulnerabilities. Who is the best editor? Who knows the agency’s history? Who understands the data? Who is experimenting effectively with new tools? 

• Modernize performance conversations. Within existing critical-element flexibilities, incorporate tool fluency, continuous learning, quality of AI-assisted work, and knowledge sharing where appropriate. 

• Protect some of the dividend. Establish a local expectation that time saved through automation will create capacity for training, process redesign, innovation, and higher-value mission work—not simply an endless increase in throughput. 

These steps begin shifting the organization from managing positions to managing capabilities and outcomes. 

Empowering Human Endeavor 

The goal of artificial intelligence in the federal government should not be to shrink the workforce or turn civil servants into passive button-pushers. It should be to remove the repetitive mechanical work that has consumed too much human capacity and redirect that capacity toward work that requires judgment, creativity, context, collaboration, and responsibility. 

The promise of AI is therefore not simply that one employee can do the work of two. It is that people—and the AI tools they command—can spend more of their time doing the work that only people can do well. That is the AI Dividend worth pursuing. 

The tools got smarter. Now the job—and the organization built around it—needs to evolve. 

What’s Next? 

Ready to explore what this looks like in practice? Explore our webinar, Architecting the Intelligence-Age Federal Workforce. We’ll dissect real-world case studies, demonstrate Dynamic Position Profile concepts, and explore visual workflow maps that leaders can begin applying Monday morning. 

About the Author 

David Doane brings three decades of technical expertise and federal service to demystify complex technology for the workforce navigating its implications daily. David delivers the clarity federal officials need to harness innovation while safeguarding against risk.