Article reader Listen + reading controls
Article reader
Preparing the reader…
Reading settings
Federal AI Talent Is a Capability System, Not a Hiring Surge¶
Calls for the federal government to recruit more artificial-intelligence talent usually begin with a real constraint: agencies cannot design, evaluate, acquire, and govern consequential AI systems without people who understand the technology deeply enough to exercise independent judgment.
The common response is a hiring surge. Direct-hire authority, pooled certificates, fellowships, special salary rates, and public-service recruiting can all help. They do not solve the problem if highly capable people enter organizations that lack product ownership, modern engineering environments, technical career paths, or durable funding.
Federal AI talent should be treated as a capability system: a set of mechanisms for obtaining expertise, organizing it around mission outcomes, developing it over time, and giving it sufficient authority to change how the institution works.
Jennifer Pahlka’s 2024 observation that government must pair mandates with enablement was well aimed. Agencies were receiving new AI responsibilities while still struggling with ordinary digital delivery. OPM subsequently issued government-wide direct-hire authority for AI roles, skills-based hiring guidance, shared talent pools, and broad AI training.1
Those are valuable components. The operating model determines whether they compound.
Begin With Work, Not Headcount¶
“We need AI talent” is too vague to support workforce planning. Agencies should map the work required across a capability lifecycle:
- mission and service design;
- product management;
- data stewardship and engineering;
- software and platform engineering;
- machine learning and applied science;
- cybersecurity and privacy;
- human factors, accessibility, and civil rights;
- test, evaluation, red teaming, and assurance;
- acquisition and vendor management;
- operations, monitoring, and incident response;
- legal, policy, and executive risk decisions.
Not every role requires a machine-learning doctorate. Many of the most important gaps involve senior technical product leaders, platform engineers, data owners, evaluators, and acquisition professionals who can integrate AI into a real service.
The workforce plan should connect each role to decisions the agency must be able to make without depending exclusively on a vendor.
Use a Portfolio of Talent Mechanisms¶
Agencies can obtain expertise through several paths:
- hire: permanent civil servants who preserve institutional knowledge and accountability;
- grow: train and develop current employees whose domain and mission knowledge are irreplaceable;
- borrow: details, fellowships, reservist models, intergovernmental personnel assignments, and temporary tours;
- buy: contractors, consultants, managed services, and research partners;
- partner: universities, FFRDCs, laboratories, nonprofits, states, and other agencies;
- automate: reduce low-value work so scarce expertise concentrates on consequential decisions.
Each mechanism has a different time horizon and risk. Contractors can scale rapidly but may accumulate mission knowledge outside the government. Fellows can introduce new practices but leave before those practices become institutional. Upskilling preserves context but requires protected time, mentorship, and real work.
The right portfolio changes by capability. The government should retain permanent authority over mission priorities, architecture, data responsibility, risk acceptance, supplier decisions, and operational accountability even when external teams do much of the implementation.
Hire Teams, Not Isolated Experts¶
An agency can recruit an excellent data scientist and produce little value if the person cannot access data, deploy software, reach users, or obtain timely security decisions. AI delivery is team work.
A minimum mission product team might include:
- a product owner accountable for the outcome;
- domain and user expertise;
- software, data, and ML engineering;
- security and privacy;
- design or human factors;
- and access to acquisition, legal, and evaluation support.
The exact composition varies, but the team needs end-to-end responsibility and a usable delivery environment. Hiring should be coordinated around these units rather than dispersed among organizational silos.
Pooled hiring can help by identifying qualified candidates once for multiple agencies. The analogous organizational move is pooled enablement: common platforms, evaluation services, legal patterns, and acquisition support that prevent every new hire from rebuilding the same foundations.
Time to Impact Matters More Than Time to Hire¶
Reducing the length of federal hiring is important. Agencies should also measure what happens after acceptance:
- time to receive equipment and access;
- time to a representative development environment;
- time to data and users;
- time to the first production or operational contribution;
- time to independent decision authority;
- and retention after one, two, and four years.
A fast offer followed by six months of onboarding friction is not a competitive talent system.
Managers should own this path. Technical onboarding needs documented architectures, development environments, code and data access, decision context, and a real initial problem. New technologists should not spend their first months discovering which undocumented relationships permit work to proceed.
Retention Depends on Professional Agency¶
Government will not always match private compensation. Mission is a genuine advantage, especially where the scale and public consequence of the work are unique. Mission cannot compensate indefinitely for preventable dysfunction.
Technical professionals remain when they can:
- solve meaningful problems;
- work with capable peers;
- use credible tools and practices;
- see their work reach users;
- exercise appropriate technical judgment;
- advance without becoming general managers;
- learn and publish where security permits;
- and receive protection when evidence challenges a favored initiative.
Agencies need technical career ladders, communities of practice, mentoring, mobility, and recognition of engineering and research leadership. A principal engineer or senior data scientist should possess status and influence comparable to a senior program manager when the mission depends on technical decisions.
Upskilling Must Be Attached to Work¶
Introductory AI literacy can help executives and staff understand terms and risks. It does not create capability by itself.
Advanced development should combine:
- role-specific curricula;
- supervised work on real agency problems;
- access to safe sandboxes and representative data;
- mentorship by experienced practitioners;
- communities that review artifacts and share patterns;
- and demonstrated competencies tied to career progression.
The agency should distinguish people who need to use approved AI tools, people who manage AI-enabled workflows, people who build systems, and people who independently evaluate them. A single training course cannot serve all four.
Domain professionals are central. The best AI team combines technical depth with people who understand policy, operations, users, failure consequences, and institutional history.
Funding Must Follow the Workforce Lifecycle¶
Hiring authority without funded positions and continuing program resources creates false capacity. AI systems require ongoing data work, evaluation, software maintenance, monitoring, security, and user support.
Congress and agencies should fund product teams and enabling platforms as continuing capabilities rather than one-time modernization projects. Workforce plans should include the full cost of supervision, tools, compute, training, and professional development—not only salaries.
Short-term funding also distorts retention. Talented people are less likely to join or stay when the organization cannot explain whether the team or product will exist after the current initiative.
Measure Capability, Not the Number of Hires¶
A federal AI workforce dashboard should move beyond offers and onboarded staff. Useful measures include:
- critical roles filled in mission product teams;
- time from hire to production contribution;
- percentage of high-impact systems with independent government technical evaluation;
- retention and internal mobility;
- diversity of professional and educational pathways;
- contractor-to-government knowledge transfer;
- reusable assets produced across agencies;
- and mission outcomes improved by teams.
These measures ask whether people changed institutional capability.
The Strategic Inference¶
The federal government does not merely compete for AI talent. It competes to offer a professional environment in which that talent can matter.
Hiring reforms are essential, especially direct hire, skills-based assessment, and pooled recruiting. Their value is realized only when people enter durable cross-functional teams, receive tools and authority, grow through meaningful work, and preserve knowledge inside the institution.
Mandates without enablement create compliance. Talent without an operating model creates frustration. The combination creates public capability.
This essay was substantially revised in July 2026 to replace the original interview summary with an evidence-based workforce-system analysis. It incorporates OPM actions published after the original post.
I write and work at the intersection of AI engineering, product strategy, and institutional transformation. I welcome public-sector technology practitioners to connect on LinkedIn.
References¶
- U.S. Office of Personnel Management, “Direct Hire Authority.”
- U.S. Office of Personnel Management, Skills-Based Hiring Guidance and Competency Model for Artificial Intelligence Work, April 29, 2024.
- Senate Committee on Homeland Security and Governmental Affairs, “Harnessing AI to Improve Government Services and Customer Experience,” January 10, 2024.
- Rebecca Heilweil, “Ex-White House official says Congress, federal agencies should do more for AI talent search,” FedScoop, January 2024. This report was the historical prompt for the original post.
-
U.S. Office of Personnel Management, “FY 2024 Human Capital Reviews—Artificial Intelligence.” ↩