Scott Kupor, now OPM Director, explains how he's applying Silicon Valley talent practices to 2.4M federal employees

Summary

Scott Kupor, former managing partner at Andreessen Horowitz, now serves as Director of the Office of Personnel Management, the federal government's central HR function overseeing 2.4 million civilian employees. His mandate is essentially a corporate turnaround: fix talent acquisition, install performance accountability, and build an AI-ready workforce before the gap becomes structural.

The Performance Problem

The core dysfunction Kupor identifies is a broken ratings distribution. Currently 70% of federal employees receive a 4 or 5 on a 5-point performance scale, while only 0.2% are rated 1 or 2. The predictable result is that top performers collect an average cash bonus of roughly 0.5% because the pool is spread too thin. Kupor wants to compress the high-rating population sharply and redirect the majority of the bonus pool to genuine top performers, a straightforward stack-ranking reform that has never been implemented at federal scale.

The incentive architecture runs deeper than bonuses. Promotion and power in the current system accrue through headcount growth, larger budgets, and tenure, not outcomes. Kupor is explicitly trying to invert that, rewarding innovation and what he calls "measured risk-taking" while acknowledging the government cannot operate with the same risk tolerance as a venture-backed startup.

Hiring Is Structurally Broken

Skills-based hiring barely exists in the federal government today. Candidates apply through USAJobs, tailor resumes to keyword-match job descriptions, and advance through a self-assessed filtering process. Kupor notes that an IT network administrator can be hired without a single technical interview. A recent legal ruling has cleared the path for skills assessments, which OPM is now moving to implement.

The sourcing model is equally passive. The government posts openings and waits. Kupor's preferred model flips that entirely: a candidate should apply once and have their profile automatically matched against all relevant openings across agencies. He estimates there are thousands of open data scientist roles across the federal government today that most qualified candidates would never find.

The recruiting pitch itself is also misaligned. The current value proposition is job stability, which Kupor flatly rejects as both dishonest and unattractive to high-agency talent. The replacement pitch centers on scale, complexity, and mission, working on genuinely hard problems with real-world consequences that no private-sector role can replicate in the same way.

A Two-Year Service Pipeline

Kupor is developing a structured short-tenure program modeled loosely on the Civilian Conservation Corps, targeting recent graduates and early-career professionals for two-year government rotations. The proposed incentives include partial student debt forgiveness and, critically, agreements with private-sector employers to guarantee jobs at the end of the rotation, with the two years counting toward seniority. The goal is to normalize a government tour as a career accelerant rather than a detour, positioning it the way Goldman Sachs or Facebook recruiting functions on college campuses today.

He is explicit that credentials should not gate this pipeline. The rejection of a candidate like Luke Farritor by unnamed government officials who cited lack of a college degree and insufficient years of experience is, in Kupor's framing, precisely the credentialism that merit-based hiring reform is designed to eliminate.

AI Strategy: Start Small, Build the Habit

Kupor currently cannot access an LLM on his government-issued computer. That baseline problem defines the starting point of OPM's AI strategy. The near-term plan is incremental: identify use cases built entirely on public data, train lightweight models, and drive 10 to 15% efficiency gains without touching confidential information or requiring airgapped infrastructure.

A working example is already in front of him. OPM recently received 40,000 public comments on a proposed merit-hiring rule, all of which staff read and responded to manually. An LLM trained on public regulatory data could materially accelerate that process with human oversight intact. Kupor is bringing in an outside consultant to map the full landscape of AI use cases across HR and customer support functions, then tasking internal teams to identify the smallest viable wins first.

The longer-term concern is workforce composition. Kupor wants to avoid arriving at an AI-transformed government in five years without the talent to run it, and sees active information-sharing and employee exchange programs with the private sector as the mechanism to stay current.

What He Found Inside

Kupor's candid assessment is that the people are better than the system deserves. Federal employees are genuinely mission-driven and take public service seriously. The failure is institutional: a structure that punishes risk, offers no upside for innovation, and has spent decades optimizing for caution over performance. His working theory is that the same people, under a redesigned incentive system, would behave very differently.