Your Federal Job Interviewer Is Now an AI: Inside the Tech Force Pilot Powered by CodeSignal

The next person grilling a candidate for a prestigious federal technology job may not be a person at all. According to a CBS News report confirmed by US officials, the federal government is moving forward with a pilot in which AI virtual agents conduct structured first-round interviews for applicants to the US Tech Force — the Trump administration’s flagship initiative to infuse agencies with roughly 1,000 elite technology specialists. The agents, built by hiring-technology firm CodeSignal, will interview candidates by phone, audio, and text, ask standardized questions, record and transcribe the answers, and pass the results to human hiring managers downstream.

On its face, it is a staffing decision at a single program. In practice, it is one of the most symbolically loaded deployments of agentic AI anywhere in government: the world’s largest employer is now using autonomous agents to help decide who gets to work on accelerating the government’s own use of AI.

What the Tech Force actually is

The US Tech Force, administered by the Office of Personnel Management (OPM), launched in December 2025 alongside a sweeping memo titled “Building the AI Workforce of the Future.” The program’s stated mission is to recruit an elite cohort of around 1,000 technology specialists — drawn heavily from the private sector — and deploy them across federal agencies to accelerate AI implementation, modernize legacy systems, and tackle problems in cybersecurity, data infrastructure, and software delivery.

The program has not struggled for attention. According to Reuters, Tech Force has attracted approximately 25,000 applicants — a ratio of roughly 25 candidates per eventual hire. That popularity is precisely what makes automation attractive to OPM: no human screening organization can cost-effectively conduct tens of thousands of structured first-round interviews in a compressed hiring timeline.

OPM’s partnership with CodeSignal already covers standardized technical skill assessments across the Tech Force hiring process, spanning software engineering, data science, and adjacent domains. The interview pilot extends that relationship from testing into the interview itself — a meaningfully different proposition, because interviews have historically been the stage where human judgment, adaptability, and candidate experience live.

How the AI interviews work

Details reported by CBS and follow-up analyses paint a picture of a tightly structured process:

• AI-led structured interviews. Virtual agents conduct first-round interviews using standardized question sets rather than free-form conversation, with the ability to ask follow-ups based on candidate responses.
• Three channels. Candidates can be interviewed by phone, by audio, or by text — an accessibility-minded design choice that lets applicants pick the mode that suits them.
• Recording and transcription. The agents record and transcribe responses, creating a written record that flows into the hiring file.
• Humans decide — formally. Hiring managers review the AI-generated summaries and transcripts and make the downstream decisions. The official framing is that the pilot “tests the machine, not the decision”: AI shapes and filters the evidence, but humans sign off.

That last point is doing a lot of work in the administration’s defense of the program, and critics are already probing it.

Twelve unanswered questions

Within days of the CBS report, watchdog site FederalHiringData published an audit identifying twelve unanswered public questions about the pilot — ranging from how AI-generated scores will be used, to what human review actually consists of, to how candidates can contest an AI’s characterization of their interview performance.

The concern is not hypothetical. Structured AI interviews are only as fair as their questions, their scoring rubrics, and their handling of accents, speech disabilities, non-native English speakers, and candidates who simply interview poorly against a machine but excel on the job. If an AI agent’s transcript or score quietly becomes the de facto decision — if hiring managers rubber-stamp the machine’s ranking because reviewing 25,000 applicants any other way is impossible — then the “human decision” safeguard becomes a formality.

There is also an accountability gap unique to government. Federal hiring operates under merit-system principles, veterans’ preference, and anti-discrimination law. When a private company’s AI screener goes wrong, the damage is one employer’s bad hire or a discrimination suit. When a federal AI interviewer systematically misjudges a category of candidates, it touches the constitutional mechanics of public employment — and no framework yet exists for auditing an LLM-based interviewer the way you audit a civil-service exam.

The irony cuts both ways

Skeptics note the deep irony: a program built to bring human technical judgment into government is using machines to filter which humans get in. Supporters reply with a different irony — that an AI-acceleration program that refused to use AI in its own operations would look worse, not better. Tech Force hires will spend their careers deploying agents into procurement, benefits processing, and citizen services; experiencing an agent-mediated hiring process is arguably on-the-job training for exactly that future.

The private sector has already moved first, and messily. Companies that rushed into one-way video interviews and automated screening over the past two years have been quietly retreating — Google, for instance, has reportedly brought back in-person interviews after finding candidates using off-camera AI tools to answer questions in real time. An arms race of AI interviewers versus AI-assisted candidates erodes the signal interviews were supposed to carry. A federal pilot entering that environment will be a stress test not just of CodeSignal’s agents but of whether structured, standardized questioning can survive when both sides of the table are augmented.

What to watch

Three things will determine whether this pilot becomes the template for federal hiring or a cautionary footnote. First, whether OPM publishes the FederalHiringData answers — score usage, human-review depth, and appeal rights — before the pilot scales. Second, whether pass rates break down along measurable demographic or linguistic lines, which outside analysts will eventually be able to detect even if OPM stays quiet. Third, whether the ~1,000 hires that emerge from the AI-screened funnel actually perform, which is the metric Tech Force was created to move in the first place.

The government has made its bet: that at a scale of 25,000 applicants, a standardized AI interview is more fair than an overloaded human one — consistent questions for everyone beats rushed questions for a lucky few. It is a defensible bet. It is also one the federal government has never before placed with technology this opaque, and the pilot’s real test is whether “trust us, a human decides” survives contact with 25,000 applicants who were judged by a machine first.