Structured interview software for government: map every rule to evidence

Structured interview software for government is only worth buying if it produces a record you can defend, because in public-sector hiring the interview is evidence, not just a decision. The right tool maps each legal obligation (merit principles, adverse impact, bias audits, retention, accessibility) to a specific artefact it creates and keeps. If a vendor can't show you that artefact, the feature doesn't count.

Key Takeaways

A public-sector interview is an evidentiary record, so judge software by the artefacts it leaves behind, not the scores it shows on screen.
OPM describes a highly structured interview as questions drawn from job analysis, the same pre-defined lead and probe questions for every applicant, and scoring against benchmarks of proficiency.
State and local governments keep employment records for two years under 29 CFR 1602.31, double the one-year period that 29 CFR 1602.14 sets for employers generally.
Accessibility is a candidate-facing obligation. The ADA Title II web rule sets WCAG 2.1 AA for state and local governments.
Ask every vendor, including us, exactly which authorizations they hold. Expert Hire does not claim FedRAMP authorization, and this piece is written for state, county, municipal and contractor buyers.

Why public-sector hiring mandates structured interviews

Government hiring runs on a legal premise private companies don't share. For federal jobs, the first merit system principle in 5 U.S.C. 2301 says selection should rest "solely on the basis of relative ability, knowledge, and skills, after fair and open competition." Most state civil service systems carry their own version of the same idea. An unstructured chat with a hiring manager is hard to square with that sentence.

The research agrees, with one update people keep missing. Schmidt and Hunter's 1998 meta-analysis put structured interview validity at .51, and that number still circulates as if it were current. Sackett, Zhang, Berry and Lievens (2022, Journal of Applied Psychology) found that earlier corrections for range restriction had overstated validity, and revised the figure to roughly .42. Structured interviews still came out as the strongest predictor they examined, and still well ahead of unstructured ones.

So the mandate is settled. What isn't settled is whether your government structured interview software can prove you followed it. Our guide to structured interview software covers the general category, and this piece is the public-sector version.

The interview is an evidentiary record, so map each obligation to an artefact

The best framing in this niche treats the interview as evidence, and it's right. But the pages built on that idea skip half the regulatory map, and the pages covering government broadly cite almost no regulation. Neither is enough for a buyer who answers to an auditor, a union, or a court.

Here's the mapping we'd use, obligation first, artefact second:

OPM guidance and merit principles map to a fixed question set tied to job analysis, and anchored rating scales written before the first candidate is seen.
EEOC adverse impact rules map to retained, per-candidate scoring data you can cut by race, sex and ethnicity.
NYC Local Law 144 maps to an independent bias audit, even for employers it doesn't bind.
The EU AI Act maps to a high-risk classification and the documentation that comes with it.
Record-keeping rules map to an export that outlives your software subscription.
Accessibility law maps to the candidate's interview flow, not the recruiter's dashboard.

If a vendor can't name the artefact for each line, you have a gap. The rest of this piece walks through them.

What OPM guidance actually asks of the question set

If you search for structured interview government guidance, the Office of Personnel Management's structured interview page comes up first, and it's worth reading in full. It describes the highest level of structure in specific terms. Questions come from competencies identified through job analysis, every applicant gets the same pre-defined lead and probe questions, and answers are scored against benchmarks of proficiency.

That last point on probes is where AI interviewers deserve scrutiny, including ours. A conversational AI that improvises follow-up questions is flexible, and flexibility is exactly what structure is meant to remove. Before you buy any federal structured interview software, ask the vendor to show you which follow-ups are pre-defined, which are generated, and how the rubric stays identical across candidates either way.

The artefacts you want here are boring and essential. You need the job-analysis document the questions came from, the question set as it was actually asked, and the rating anchors written before scoring began. Our explainer on how AI interviews are scored walks through what an anchored, per-criterion score should look like.

In Expert Hire, a hiring manager can edit the rubric before any interview runs. Calibration mode lets you run a known candidate through and tune the rubric until the score matches your expectation. Each score then carries the rubric criterion, the transcript excerpt, and the AI's written reasoning. That's the shape of evidence OPM's benchmark language points at, and our methodology page documents the research behind the rubric design.

Adverse impact, bias audits, and the scoring data you have to keep

The federal Uniform Guidelines are blunt about records. 29 CFR 1607.4 says each user should keep records that disclose the impact of its selection procedures by race, sex and ethnic group. The same section sets the four-fifths rule: a selection rate for any group below 80% of the highest group's rate is generally treated as evidence of adverse impact.

You can't run that analysis on a gut feeling logged as "strong yes." You need numeric scores, per criterion, tied to candidate records and retained long enough to aggregate. That's the practical case for anchored rubrics, and it's also why we'd argue adverse-impact analysis is part of validating your process, not a separate compliance chore. For the upstream work, our piece on reducing bias in hiring covers where bias enters before the interview even starts.

Bias audits are the next layer. New York City's Local Law 144 requires an automated employment decision tool to have had a bias audit by an independent auditor within one year of its use, public posting of the results, and notice to candidates.

A county in Ohio isn't bound by it. But its rules spell out what the audit has to report (selection rates and impact ratios by sex and race or ethnicity). That makes it a useful spec to write into an RFP.

A posted audit isn't proof on its own. In December 2025 the New York State Comptroller audited the city's enforcement of the law and called it ineffective. It cited complaints that never reached the enforcing agency and reviews of posted audits that missed potential violations the Comptroller's own auditors found.

Expert Hire provides Local Law 144 AEDT-ready bias audit documentation, but audit-ready documentation isn't the same thing as an independent audit. So ask every vendor, us included, who the independent auditor was and when the last audit ran.

Records retention: two years for public employers, and outliving your vendor

The one vendor guide that goes deep on retention quotes a one-year federal floor, which comes from 29 CFR 1602.14, the general rule for employers. State and local governments fall under a different section. 29 CFR 1602.31 requires a political jurisdiction to keep personnel and employment records for two years from the making of the record or the personnel action, whichever is later.

Both rules stretch further once a charge or lawsuit is filed, because relevant records must then be kept until final disposition. A discrimination case can easily outlast a three-year software contract. If your interview records live only inside a vendor's platform, your retention obligation now depends on a procurement decision someone makes in 2028.

So the artefact here is an export, and it should land in a system you control. Expert Hire exports the shortlist to your ATS with the full scorecard attached, and scorecards sync back to candidate records in systems like Workday, Oracle HCM and SAP SuccessFactors. Whatever you buy, test the export before you sign, and check that it carries scores, transcript excerpts and reasoning, not just a pass or fail flag.

Consent sits alongside retention. Illinois requires candidate consent before AI analysis of video interviews, and Expert Hire has a built-in consent flow for the Illinois AI Video Interview Act. For the wider regulatory picture across states, see our breakdown of AI hiring laws rather than a restatement here.

Accessibility is a candidate-facing requirement, not a dashboard feature

It's easy for an accessibility claim in public sector hiring software to cover only the admin console. That's the wrong surface. The person the law protects is the applicant taking the interview, often on a phone, sometimes with a screen reader, and occasionally with an accommodation request that has to be honoured mid-process.

For federal agencies, the Revised Section 508 Standards incorporate WCAG 2.0 Level AA for ICT that agencies procure or use. For state and local governments, the Department of Justice's ADA Title II web rule sets WCAG 2.1 Level AA. After an interim final rule extended the dates, entities with 50,000 or more people must comply by April 26, 2027, and smaller entities by April 26, 2028.

Ask for an Accessibility Conformance Report that covers the candidate interview flow specifically, then test it yourself with a keyboard and a screen reader. Also ask how an accommodation (extra time, a different format, a human interviewer) is recorded, because that decision is part of the evidentiary record too.

What to verify before buying structured interview software for government

The EU AI Act belongs on this checklist if any of your hiring touches the EU, and it's a useful benchmark even if it doesn't. Annex III, point 4 classifies AI systems used to evaluate candidates as high-risk. Under Article 27, deployers that are bodies governed by public law must carry out a fundamental rights impact assessment before deploying one. Expert Hire provides EU AI Act risk-classification documentation per use case.

Then there's authorization, where honesty matters most. At least one enterprise vendor in this space holds FedRAMP authorization, and federal agencies will ask everyone else for it. Expert Hire does not claim FedRAMP authorization. Our SOC 2 and ISO 27001 work is published as a roadmap, so ask us for the current state in writing rather than assuming an attestation.

So we'd point federal agencies elsewhere today and focus on state, county and municipal teams, plus federal contractors.

Before you sign any government structured interview software contract, get written answers to these:

Question structure: which lead and probe questions are fixed, and which are generated?
Scoring evidence: does every score carry the criterion, the transcript excerpt and the reasoning?
Adverse impact: can you pull per-candidate scores by group without a vendor ticket?
Bias audit: who was the independent auditor, and when was the last audit?
Retention and export: does the export survive contract termination in a system you own?
Accessibility: does the conformance report cover the candidate flow, tested against WCAG 2.1 AA?
Authorization: which certifications are held today, and which are roadmap?

If the role is technical (state IT, data teams, engineering at a transit authority), also check whether the tool can assess the actual skill rather than a proxy. Expert Hire runs a live, conversational AI interview with a live coding environment for software, data and ML roles. Human judgement still owns the final call, which is why we don't think AI will replace recruiters.

Frequently asked questions

What is structured interview software for government?

It runs or supports interviews where every candidate gets the same job-related questions, scored against the same predefined rating scales. In government it also has to keep the record: questions asked, scores, reasoning, and the data needed for adverse-impact analysis.

Is federal structured interview software different from state or local tools?

The interview method is the same, but the obligations differ. Federal agencies work under Title 5 merit principles, Section 508 and, for cloud services, FedRAMP. State and local governments work under civil service rules, the ADA Title II web rule and a two-year record-retention requirement under 29 CFR 1602.31.

Do AI interviews count as structured interviews?

They can, if the questions, probes and rubric are fixed in advance and applied identically. An AI that improvises different follow-ups for different candidates weakens the structure, so ask vendors to show exactly what is fixed and what is generated.

How long do public employers have to keep interview records?

Under 29 CFR 1602.31, political jurisdictions keep employment records for two years from the record or the personnel action, whichever is later. If a charge or lawsuit is filed, relevant records must be kept until it's resolved.

Does NYC Local Law 144 apply to government employers outside New York City?

No, it applies to employers and employment agencies using automated tools for jobs in New York City. It's still worth borrowing its bias-audit requirements as a procurement standard, because its rules spell out exactly what an audit has to report.

Conclusion: buy the record, not the interview

In public-sector hiring the interview is an evidentiary record, and the software you buy is really the system that creates and keeps that record. Map every obligation to the artefact that satisfies it, test the candidate flow and the export yourself, and treat any vendor who blurs what they're authorized for as a risk.

If your team hires for technical roles at the state, county or city level, see how Expert Hire's AI interview platform produces a scored report card with the transcript behind every score. Then check it against the checklist above.

By TK, Growth at Expert Hire. Last updated September 29, 2026. Reviewed by Anand Suresh, CPO at Expert Hire.