TL;DR: Skills-based hiring judges demonstrated, job-relevant skills rather than degrees or job titles. In our framing, a structured interview scored against anchored criteria is a skills method; among the methods the research compares, it has the highest operational validity, .42 in Sackett et al. (2022). The honest limit is that .42 is for human-rated interviews; no meta-analysis exists for AI-scored ones.
Updated 21 October 2026 · What changed: first publication.
Key takeaways
- Two surveys give two answers on adoption: NACE says 70% (n=183) and TestGorilla says 85% (n=1,084 decision-makers). Sampling frame, the definition of "uses" and vendor interest explain the gap.
- Announcements have outrun practice. The Burning Glass Institute and Harvard Business School found that not even 1 in 700 hires were affected by dropped degree requirements (February 2024). That is why skills-based hiring needs process evidence.
- Structured interviews have the highest operational validity of the common methods in Sackett et al. (2022): .42, against job knowledge tests at .40, work samples at .33 and cognitive ability at .31.
- The Black-White standardized difference is .23 for structured interviews and .54 to .79 for knowledge, work-sample and cognitive tests. That is one input to a method choice, not a guarantee of fairness.
- For AI-scored interviews, no meta-analytic validity estimate exists. What AI can responsibly add is evidence for a human to check.
Depending on the survey, 70% or 85% of employers say they hire on skills. Yet the Burning Glass Institute and Harvard Business School study found that dropping degree requirements affected not even 1 hire in 700. The gap is a process problem: a skill you never measure is a claim, not a criterion. This page shows how a structured interview works as a skills method, which other methods to weigh it against, and where AI helps and where it does not. AI interviews are one way to generate per-skill evidence at volume, and how AI interview software works is the place to start if you are new to the category.
What is skills-based hiring?
Definition. Skills-based hiring evaluates candidates on demonstrated, job-relevant skills rather than degrees or job titles. The employer defines the skills a role needs, assesses each one with evidence, such as a work sample, a test or a scored structured interview, and keeps the record that supports the decision.
How it differs from the traditional approach:
| Traditional hiring | Skills-based hiring | |
|---|---|---|
| Basis of the screen | Degree, job title, years of experience | Demonstrated, job-relevant skills |
| Job description | Lists credentials | Lists the skills the job needs |
| Interview | Open conversation, impressions | Same questions, scored against anchors |
| Evidence kept | Notes, a gut call | Per-skill ratings and records |
Skills-based and competency-based hiring overlap. In our usage, a competency bundles skills with knowledge and behaviours, while a skill is one specific, demonstrable ability. The method on this page works for both.
How many employers actually use skills-based hiring?
Two surveys give two answers, and both are self-reported.
NACE Job Outlook 2026 reports 70% of responding employers use skills-based hiring, up from 65% a year earlier; 71% of those use it at least half the time. By stage, 87% use it in interviewing, 81% in job descriptions, 65% in screening and 58% in interview rubrics. GPA screening fell from 73% in 2019 to 42%. The survey had 183 respondents (170 NACE employer members and 13 non-members), was fielded 7 August to 22 September 2025 and was published 12 January 2026. It is member-skewed toward college-hiring employers. (NACE)
TestGorilla's State of Skills-Based Hiring 2025 reports 85% of employers use it, up from 81% in 2024 and 73% in 2023, and 76% use skills tests. It surveyed 1,084 hiring decision-makers and 1,076 job seekers in the UK and US in April 2025 and was published 6 June 2025. It is a skills-testing vendor's own survey. (TestGorilla)
Why they differ: the sampling frame (college-hiring employers against a vendor's panel), the definition of "uses", the respondent's role, and vendor interest. Pick the one whose sample looks like your market, and treat neither as a settled number. HireVue's own 2025 AI report adds that 41% of HR professionals say their company uses skill assessments, from a vendor survey published 19 February 2025 (HireVue). The broader sourced set of adoption and candidate-trust figures is on the AI recruitment statistics page.
Does skills-based hiring change who gets hired?
Less than the announcements suggest. The Burning Glass Institute and Harvard Business School study, Skills-Based Hiring: The Long Road from Pronouncements to Practice (Sigelman, Fuller and Martin, February 2024), found that not even 1 in 700 hires were affected by dropped degree requirements. Roughly 37% of firms made real changes, about 45% changed in name only and about 20% backslid. Where firms led, outcomes were better: at leader firms, retention of non-degreed hires was 10 points above their degreed colleagues, and non-degreed workers hired into formerly degree-required roles saw a 25% average salary gain. (Burning Glass Institute; summary)
Announcements outran practice, so "skills-based" needs process evidence: scorecards, rubrics and records. That is where structured interviews come in.
How do you assess skills without a test battery?
Use a structured interview. In the meta-analysis by Sackett, Zhang, Berry and Lievens (2022, Journal of Applied Psychology 107(11), 2040-2068), it has the highest operational validity of the common methods. Operational validity is the estimated correlation with job performance. The values below are drawn from the same study.
| Method | Operational validity | Black-White d |
|---|---|---|
| Structured interview | .42 (80% credibility lower bound .18) | .23 |
| Job knowledge test | .40 | .54 |
| Work sample | .33 | .67 |
| Cognitive ability test | .31 | .79 |
| Unstructured interview | .19 | Not stated |
The spread is large: the structured interview's 80% credibility lower bound is .18, so a poorly run structured interview can land well below .42. These are human-rated interviews. SIOP's The Industrial-Organizational Psychologist summarised the paper by saying structured interviews "may in fact be the strongest predictor of job performance—not cognitive ability" (SIOP TIP). The full table, with the 1998 comparison, is in structured vs unstructured interview validity.
Which skills-assessment method should you use?
Match the method to the skill and the volume. The table sets validity and subgroup difference beside the practical trade-offs, and marks a cell "Not stated" where the data does not exist. The validity and d columns are from Sackett et al. (2022). The last four columns are our judgement, not sourced figures. We name tool categories, not vendors, because the vendor rankings we found are self-ranked.
| Method | What it measures | Operational validity | Black-White d | Candidate time | AI-assist exposure | Evidence artefact | Best used when |
|---|---|---|---|---|---|---|---|
| Work sample | Performance on a sample of the job's tasks | .33 | .67 | High | Higher if unsupervised | Scored output | The task is concrete and time allows |
| Job knowledge test | Knowledge the job requires | .40 | .54 | Medium | Higher if unsupervised | Test score | Knowledge is the core requirement |
| Cognitive ability test | General reasoning | .31 | .79 | Medium | Higher if unsupervised | Test score | Role learning demands are high; weigh adverse impact |
| Structured interview (human) | Answers to job-related questions, scored on anchors | .42 (.18) | .23 | Medium | Medium | Rubric rating, notes | You need a record of each skill |
| Structured interview (AI-delivered, rubric-scored) | The same, at volume | No meta-analytic estimate exists for AI-scored interviews; the .42 figure is for human-rated structured interviews | Not stated | Medium | Medium | Transcript, per-answer score, reasoning | You need volume with a human decision |
| Résumé skills parsing | Skills text extracted from CVs | No comparable estimate in Sackett | Not stated | Low | Higher | Extracted skill list | A first filter, never the final evidence |
One legal reminder. Skills tests are selection procedures and can show adverse impact, and the EEOC's Uniform Guidelines apply.
Can an AI interview assess skills?
As an evidence generator against a rubric, yes. As a validated skills test, no.
What AI adds: the same questions to everyone, a per-answer record, and scoring against a stated answer key. What it does not add: a validity estimate. The .42 above is for human-rated interviews, and no meta-analysis exists for AI-scored ones.
The candidate side calls for care. Gartner found that 26% of candidates trust AI to evaluate them fairly (n=2,918, 31 July 2025), and Greenhouse found that 70% were never clearly told AI would evaluate them (n=2,950, 1 May 2026). (Gartner; Greenhouse) Test-based assessment has its own exposure: Gartner reports 6% of candidates admitted interview fraud (2Q25, n=3,000) and 39% used AI in the application process (4Q24, n=3,290), per the same Gartner release. That supports pairing skills evidence with verification, and we make no identity or fraud-detection claim here.
According to TestGorilla's 7 October 2025 announcement, the survey publisher has launched skills-based AI video interviews in one-way and conversational formats, scored against structured, skills-based criteria (BusinessWire). In our reading, if that shape (structured questions, an explainable score) is becoming standard, the differentiator is the human decision gate and operating at staffing scale.
What this looks like in an interview platform
In ZeaHire, jobs carry Required Skills tags, and interview questions can be AI-generated by count, difficulty, industry and skills. Each answer is scored with the AI's reasoning visible. Resumes receive Skills, Experience and Education sub-scores, and a person approves or rejects. That is a per-question evidence record for a human to weigh. It is not a validated skills test, and no validation study of ZeaHire's scores is available.
Explanation matters as much as the score. Each score should come with the answer, the key and the reasoning, which is the point of explainable AI scoring.
For readers ready to evaluate tools, choosing AI interview software sets out the criteria.
How do you implement skills-based hiring in six steps?
- Define the skills from the job, not the previous holder. Artefact: a skills list tied to tasks.
- Write behavioural or situational questions for each skill. The structured interview question bank supplies question text and rubrics. Artefact: a question set.
- Anchor a rubric. Describe 1, 3 and 5 in content terms, as the bank does. Artefact: a scoring guide.
- Score per skill. Rate each skill separately before you form an overall view. This is practice guidance for any process, not a description of what any one tool reports. Artefact: a skill-by-skill scorecard.
- Record the evidence. Keep the answer, the rating and the reason together. Artefact: an auditable record.
- Make the decision with a person, and document the reasons. Artefact: a decision note.
Skills-based hiring for staffing agencies
When you submit candidates to clients, per-skill evidence is what lets you show why a candidate fits rather than assert it. Keep the same skills record for each submission and map it to each client's requirements. The staffing detail sits in AI interviews for staffing agencies.
Frequently asked questions
What is skills-based hiring?
Skills-based hiring evaluates candidates on demonstrated, job-relevant skills rather than degrees or job titles. The employer defines the skills, assesses each with evidence and keeps the record behind the decision.
Does skills-based hiring work?
Where firms made real changes, outcomes were better, but most announcements did not change hiring. The Burning Glass Institute and Harvard Business School found retention of non-degreed hires 10 points higher at leader firms, and a 25% average salary gain, but not even 1 in 700 hires were affected overall.
How do you assess skills without a test?
Use a structured interview scored against a rubric. Sackett et al. (2022) put the operational validity of structured interviews at .42, with the caveat that this is for human-rated interviews.
Are skills tests biased?
Every selection procedure can show adverse impact, the EEOC's Uniform Guidelines apply to tests and interviews alike, and the differences vary by method. In Sackett et al. (2022) the Black-White standardized difference is .23 for structured interviews and .54 to .79 for knowledge, work-sample and cognitive tests. That does not make structured interviews unbiased, so monitor outcomes.
Can an AI interview assess skills?
Yes, as an evidence generator against a rubric, not as a validated test. No meta-analysis exists for AI-scored interviews.
What is the difference between skills-based and competency-based hiring?
A skill is a specific, demonstrable ability; a competency usually bundles skills with knowledge and behaviours. The two overlap in practice.
Do you still need a CV?
A CV remains a source of claims to verify, and skills evidence supplements it. The CV tells you where to look; the interview and the work sample show what the person can do.
How many employers actually do it?
NACE reports 70% (n=183) and TestGorilla reports 85% (n=1,084 decision-makers). The gap reflects sampling, the definition of "uses" and vendor interest.
For a current example of a large AI company entering hiring, see where OpenAI's hiring platform stands.
Next step
Book a demo to see per-question evidence, with the reasoning behind each score.
Comments
No comments yet — start the conversation.