Classet
Classet's AI voice recruiter, Joy, phones hourly and frontline applicants within about a minute of applying, runs employer-scripted screening interviews with binary knockout questions, and sends recruiters recordings, transcripts, summaries and a qualified/unqualified outcome, either in Classet's lightweight ATS or in 100+ connected ATSs.
HireAIScore rates Classet 55 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed September 25, 2026. Classet ranks 27 of 160 vendors rated in Screening AI, against a category average of 46. That score is drawn from 38 evidence items across 7 rubric criteria for Classet — 33 with a cited source, 5 recording that nothing was located.
§ 01 - Company facts
- Legal name
- Classet, Inc.
- Founded
- 2021
- Headquarters
- United States · US
- Pricing tier
- Mid
- Market side
- Employer-side (AEDT)
- Categories
- ScreeningATS-integratedScheduling
- Website
- classet.ai
These are company details, not findings. They are editable by a verified representative and carry no weight in the score, which is built only from the cited evidence below.
§ 02 - Score breakdown
§ Score breakdown
Category scoring
Weighted contribution shown to the right of each bar.
- 01
Article 11 Technical Documentation
Weight 20%46
+9.2 · category avg 49
- 02
Bias Audit Transparency
Weight 18%76
+13.7 · category avg 41
- 03
FRIA Support
Weight 15%33
+5.0 · category avg 31
- 04
Data Governance Disclosure
Weight 15%57
+8.5 · category avg 53
- 05
Human Oversight Design
Weight 12%55
+6.6 · category avg 56
- 06
Post-Market Monitoring
Weight 12%58
+7.0 · category avg 40
- 07
Customer Documentation
Weight 8%62
+5.0 · category avg 58
Category avg is the mean raw score on that criterion across the 160 Screening AI vendors in scope of this rubric, this one included.
§ 03 - Strongest · weakest
Strongest category
Bias Audit Transparency
Raw score 76 · contributes 13.7 to total.
76 against a 41 category average
Weakest category
FRIA Support
Raw score 33 · contributes 5.0 to total.
33 against a 31 category average
§ 04 - Cited evidence
Download diligence record→§ Evidence
Cited per category
Every score is backed by at least one cited piece of evidence.
Evidence ledger
- Items
- 38
- Documentation
- 18
- Audit report
- 3
- Public statement
- 12
- Absence
- 5
- With a source URL
- 33 of 38
- Source hostnames
- 5
- Fewest items
- 4
- Bias Audit Transparency
These figures measure how thoroughly Classet was reviewed, not how Classet performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
5 items46
- DocumentationCaptured Sep 25, 2026
Classet's public API docs define each interview outcome, including 'qualified' and 'unqualified' (the candidate passed or failed employer-configured qualifying questions), 'ai_opt_out' and 'ai_stopped_override'.
https://docs.classet.ai/reference/frequently-asked-questions
- DocumentationCaptured Sep 25, 2026
The call-retrieval API schema returns a qualified/unqualified outcome_status with an outcome_reason, per-question fields including a boolean score_answer, a dealbreaker flag and an anticipated_answer, and AI call-analysis fields that include a candidate sentiment label.
- Public statementCaptured Sep 25, 2026
An August 2026 explainer for legal reviewers describes what Joy does (structured interviews, employer-defined knockouts, recordings, transcripts and summaries) and what Classet says it never does (rank candidates, score tone, accent or personality, analyze video, or infer individual demographics).
- DocumentationCaptured Sep 25, 2026
Classet's trust center lists an 'AI Ethics and Governance Policy', but like all 20 of its documents it is available only on request.
- AbsenceCaptured Sep 25, 2026
A crawl of about 250 classet.ai pages, the docs.classet.ai index and the trust center found no model or system card, instructions for use, ISO/IEC 42001 certification or NIST AI RMF mapping, and Classet's pages disagree on whether Joy produces scores.
Bias Audit Transparency
4 items76
- Audit reportCaptured Sep 25, 2026
Warden AI's public dashboard for Classet's AI Phone Interviewer shows a monthly audit cadence, a latest audit on 1 September 2026 covering 903,555 records, and 14 Clear, 0 Consider and 0 Concern results across sex, race/ethnicity and intersectional checks, with NYC Local Law 144 the only regulation mapped.
- Audit reportCaptured Sep 25, 2026
The freely downloadable NYC LL 144 report, dated 1 September 2026, uses historical data on real candidates and reports impact ratios of 0.96 to 1.00 (lowest: Hispanic candidates and Hispanic women, at 0.96); its PDF counts a 'Yes' match label as selection and says demographics were inferred from name- or text-based signals.
https://trust.warden-ai.com/classet/ai-phone-interviewer/nyc
- Public statementCaptured Sep 25, 2026
Classet says the audit estimates demographics in aggregate through name-based inference and Bayesian Improved Surname Geocoding, and never attaches them to individual candidates.
- Public statementCaptured Sep 25, 2026
An August 2026 Classet post claims that because nothing autonomously scores or ranks, Joy 'sits outside AEDT classification' under NYC Local Law 144.
https://www.classet.ai/blog/hyring-reviews-pricing-alternatives
FRIA Support
4 items33
- AbsenceCaptured Sep 25, 2026
Searches of classet.ai (sitemap crawl including the privacy notice and terms), docs.classet.ai and the trust center found no Fundamental Rights Impact Assessment template, Article 26/27 deployer guidance or EU instructions for use, and the Warden dashboard has no EU AI Act report.
- Public statementCaptured Sep 25, 2026
A July 2026 guide sets out employers' duties under Colorado SB 26-189 (pre-use notice, a plain-language explanation and human review within 30 days of an adverse decision, three-year record retention) and says a human-in-the-loop design does not exempt them.
- Public statementCaptured Sep 25, 2026
A May 2026 buyer's guide summarizes EU AI Act deployer duties but gives superseded dates, saying Colorado SB 24-205 took effect on 1 February 2026 and that EU high-risk obligations start on 2 August 2026.
https://www.classet.ai/blog/ai-recruiting-bias-audits-compliant-platform
- DocumentationCaptured Sep 25, 2026
The trust center offers deployers a request-only package of 19 policies, including Risk Management and AI Ethics and Governance, plus the SOC 2 audit report.
Data Governance Disclosure
5 items57
- DocumentationCaptured Sep 25, 2026
The Oneleet-hosted trust center shows SOC 2 Type 2 as compliant and GDPR as in progress, and names 40 subprocessors (21 flagged as handling personal data) with purpose descriptions, including Retell AI, which stores interview audio recordings and transcripts.
- DocumentationCaptured Sep 25, 2026
The privacy notice (updated 15 September 2026) lists voice recordings, optional demographic/EEO data, biometric data, 'sentiment analysis results' and AI inferences such as 'assessment scores' and 'candidate fit analysis', says interview responses and outcomes are used to refine Classet's AI systems, and gives only a generic 'as long as necessary' retention rule.
- Public statementCaptured Sep 25, 2026
Classet says Joy does not score tone, accent or personality or analyze video, that inferred demographics stay aggregate, and that it shares data retention and deletion terms in its review package rather than publishing them.
- DocumentationCaptured Sep 25, 2026
The healthcare page says Joy does not collect protected health information and that customers can configure how long interview data is kept to match their own policies.
- AbsenceCaptured Sep 25, 2026
No specific retention period for interview recordings or transcripts and no public DPA were found in the privacy notice, the terms of service, the API docs or the trust center, where the Data Management and Retention Policy is request-only.
Human Oversight Design
8 items55
- DocumentationCaptured Sep 25, 2026
API outcomes include 'ai_opt_out' (the candidate asked to speak with a human instead), 'ai_stopped_override' (a recruiter or admin stopped the interview) and 'unqualified' (the candidate failed an employer-set qualifying question).
https://docs.classet.ai/reference/frequently-asked-questions
- Public statementCaptured Sep 25, 2026
Classet says it discloses the AI in text, email and phone interactions, and that candidates can opt out at any time, with opt-outs routed to a designated ATS stage for human follow-up.
- Public statementCaptured Sep 25, 2026
A January 2026 product update lets employers continue an interview after a candidate misses a hard requirement instead of rejecting automatically, and says recruiters can still review candidates who miss requirements.
- Public statementCaptured Sep 25, 2026
Classet's glossary says knockout questions 'auto-qualify or auto-disqualify' candidates and that those who fail them 'don't get a recruiter's time'.
- DocumentationCaptured Sep 25, 2026
The ATS page says Joy screens applicants against must-haves and 'moves the best candidates on to autoscheduled next-round conversations'.
- Public statementCaptured Sep 25, 2026
The July 2026 Colorado post says Joy rejects no one and that, when a candidate misses a must-have criterion, it 'completes the full interview anyway and flags the miss for the recruiter'.
- DocumentationCaptured Sep 25, 2026
The Starter plan lists 'Summaries, scores, and transcripts' among its features.
- DocumentationCaptured Sep 25, 2026
The homepage product mock-up shows 'AI-powered interview results and recommendations' with candidate statuses such as 'Shortlisted' and 'Consider'.
Post-Market Monitoring
6 items58
- Audit reportCaptured Sep 25, 2026
Warden AI's dashboard presents monthly independent bias auditing of the AI Phone Interviewer, though only one dated evaluation (1 September 2026) is visible.
- DocumentationCaptured Sep 25, 2026
A public Better Stack status page tracks eight components, including Telephony and Web Calls, with 90-day uptime history and short auto-detected outages, and offers email, RSS, JSON, webhook and Slack subscriptions, but publishes no incident reports or postmortems.
- DocumentationCaptured Sep 25, 2026
A public product changelog on docs.classet.ai runs to 35 pages of releases and hotfixes, including AI model changes.
- DocumentationCaptured Sep 25, 2026
One changelog entry discloses a fix that isolated recruiter call transcripts to each organization, closing a gap that 'could have exposed other companies' interview data'.
https://docs.classet.ai/changelog/fixed-security-patch-recruiter-call-transcripts
- DocumentationCaptured Sep 25, 2026
The trust center names security@classet.org as its contact and lists controls for incident response, breach notification and penetration testing within the last 12 months, while the Incident Response Policy itself is request-only.
- AbsenceCaptured Sep 25, 2026
No security.txt (www.classet.ai/.well-known/security.txt returns 404) or public vulnerability-disclosure policy was found, and no Atlassian Statuspage exists at classet.statuspage.io.
Customer Documentation
6 items62
- DocumentationCaptured Sep 25, 2026
Pricing is published: Starter costs $249 a month billed annually ($349 month to month) for 50 interviews, Scale costs $999 a month, and Enterprise with ATS Sync starts at $1,995 a month.
- DocumentationCaptured Sep 25, 2026
The public API docs walk deployers through the candidate journey, including outreach that names 'our AI Recruiting Assistant', a choice between web and phone calls, and a reminder cadence that stops after the fourth nudge.
- Public statementCaptured Sep 25, 2026
A review guide for legal, security and compliance teams explains Joy's role and lists the materials Classet provides in reviews, including Warden results, the SOC 2 report, a knockout-configuration walkthrough, sample artifacts, retention and deletion terms, and its position on candidate notice.
- Public statementCaptured Sep 25, 2026
A July 2026 guide explains Colorado's SB 26-189 rewrite and the notice, explanation and record-retention duties it leaves employers.
- DocumentationCaptured Sep 25, 2026
The terms (last updated 23 April 2025) name NYC Local Law 144 and Illinois' AI Video Interview Act but leave compliance with employment laws to customers, and they say nothing about TCPA consent for calls or texts.
- AbsenceCaptured Sep 25, 2026
No public DPA, NYC candidate-notice template, EU deployer instructions or TCPA consent guidance was found across classet.ai, docs.classet.ai and the trust center; Classet says it shares retention terms and its candidate-notice position only during sales reviews.
§ 05 - Editorial notes
Company overview
Classet, Inc. is registered in Illinois and was founded in 2021 by the team behind Bluecrew, a staffing app later bought by EmployBridge. LinkedIn lists its headquarters as Chicago. Its product is Joy, a phone-first AI recruiter for high-volume hourly hiring in skilled trades, warehouses, healthcare, data centers and staffing. Joy calls applicants within about 60 seconds of applying and runs an employer-scripted interview with knockout questions in 25+ languages. It follows up by SMS and email, then returns a recording, transcript and summary with a qualified or unqualified outcome. Classet sells Joy three ways: with its own lightweight 'self-driving' ATS, as a plugin that syncs with 100+ ATSs, or through an API. Published prices run from $249 a month to $1,995 a month and up. The platform relies on Retell AI, which stores interview audio and transcripts, and on models from OpenAI, Google, Microsoft Azure and Anthropic.
Regulatory exposure
Joy turns employer-set knockout answers into a qualified/unqualified outcome for high-volume US hiring, so NYC Local Law 144 is its most direct exposure. The law has been enforced since July 2023, and DCWP committed to tougher enforcement after a December 2025 New York State Comptroller audit found its enforcement ineffective. Classet commissions a monthly Warden AI audit for the law, yet one of its blog posts says Joy sits outside AEDT classification, so NYC employers should decide the question themselves and give candidates notice. Illinois HB 3773 has required notice when AI is used in employment decisions since 1 January 2026. Classet's privacy notice lists biometric data among the data it processes, even though its marketing says voice screening falls outside BIPA. Colorado's SB 26-189 (effective 1 January 2027) and California's FEHA automated-decision rules (in force since October 2025, with four-year record retention) add notice, explanation and record-keeping duties, and California's CPPA ADMT rules follow on 1 January 2027. In the EU, recruitment AI is high-risk under Annex III, with those obligations now due on 2 December 2027. The ban on workplace emotion recognition has applied since February 2025, so the candidate 'sentiment' label in Classet's API would need review before any EU use. Outbound AI-voice calls and texts to applicants also raise TCPA consent questions that Classet's terms and privacy notice never address.
Path to a higher score
Classet would gain most by reconciling its own descriptions of the product. It should publish one system card that says plainly whether Joy scores, auto-disqualifies or auto-advances candidates, and then align the pricing page ('scores'), the ATS page's promise of auto-scheduled next rounds, the API's qualified/unqualified outcome and Warden's 'does not score, classify, or rank' description. The Warden audit should cover protected characteristics beyond sex and race and add EU, Colorado, California and Illinois reports. It should also report unknown-category counts and the data window, and keep a visible month-by-month history. Documentation and data-governance scores would rise if Classet published its AI Ethics and Governance policy, a DPA, fixed retention periods for recordings and transcripts, TCPA consent guidance and NYC candidate-notice templates. EU deployers would also need FRIA and deployer guidance and an explanation of the sentiment output. Incident write-ups on status.classet.org, a security.txt or vulnerability-disclosure policy, and corrected regulatory dates in its May 2026 compliance guide would round out the gains.
§ Regulatory frame
What applies to screening ai.
Treated as high-risk under Annex III §4 when the screening output is decisive. EU AI Act Article 13 transparency obligations apply; deployers must give candidates a way to contest.
§ Compare
Build any comparison→Classet against its nearest-scoring peers.
In Screening AI.
§ Others rated in Screening AI
All screening vendors→Ranked 27 of 160 by weighted total under rubric v1.0. The ordering is arithmetic on the rubric and carries no view on which tool suits a given hiring process.
- 24HackerRank55F
- 25Gem55F
- 26iCIMS55F
- 27ClassetThis profile55F
- 28Plum54F
- 29Metaview54F
Conflicts of interest
No vendor pays for placement, scoring, or removal. Casework - the consulting firm that operates this directory - provides paid services to some vendors. Any active or recent (within 24 months) commercial relationship is disclosed on the affected vendor profile and the review is reassigned to an independent reviewer. See the full policy on About.
Casework has no commercial relationship with this vendor.