Juicebox (PeopleGPT)
HireAIScore rates Juicebox (PeopleGPT) 57 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed June 7, 2026. Juicebox (PeopleGPT) ranks 11 of 86 vendors rated in Candidate sourcing AI, against a category average of 46. That score is drawn from 16 evidence items across 7 rubric criteria for Juicebox (PeopleGPT) — 12 with a cited source, 4 recording that nothing was located.
Score breakdown
Juicebox (PeopleGPT), criterion by criterion
Each rubric criterion is scored out of 100 on the cited evidence, then weighted into the total.
Weighted contribution shown to the right of each bar.
Article 11 Technical Documentation
Weight 20% · category avg 48
52
+10.4
Bias Audit Transparency
Weight 18% · category avg 40
84
+15.1
FRIA Support
Weight 15% · category avg 31
35
+5.3
Data Governance Disclosure
Weight 15% · category avg 53
56
+8.4
Human Oversight Design
Weight 12% · category avg 56
50
+6.0
Post-Market Monitoring
Weight 12% · category avg 40
55
+6.6
Customer Documentation
Weight 8% · category avg 57
60
+4.8
Total weighted score
56.6(rounded to 57 for display)
Category avg is the mean raw score on that criterion across the 86 Candidate sourcing AI vendors in scope of this rubric, this one included.
Strongest criterion
Bias Audit Transparency
84raw / 100
Contributes 15.1 to the total. 84 against a 40 category average.
Weakest criterion
FRIA Support
35raw / 100
Contributes 5.3 to the total. 35 against a 31 category average.
Cited evidence
The evidence behind the score
Every score is backed by at least one cited piece of evidence. Open a criterion to see what was found, when, and where.
Download diligence recordEvidence ledger
- Items
- 16
- Documentation
- 8
- Audit report
- 3
- Public statement
- 1
- Absence
- 4
- With a source URL
- 12 of 16
- Source hostnames
- 6
- Fewest items
- 2
- Article 11 Technical Documentation
These figures measure how thoroughly Juicebox (PeopleGPT) was reviewed, not how Juicebox (PeopleGPT) performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
2 items52
- DocumentationCaptured Jun 7, 2026
Warden AI's public LL144 report documents the Autopilot system, its inputs/outputs (job criteria and LinkedIn profiles in, 0-100 fit score out), and the disparate-impact methodology, serving as partial technical/explainability documentation.
- AbsenceCaptured Jun 7, 2026
No model card, system card, ISO/IEC 42001 certification, or standalone explainability/AI policy whitepaper was found on juicebox.ai, the docs.juicebox.work site, or the trust center.
Bias Audit Transparency
3 items84
- Audit reportCaptured Jun 7, 2026
Publicly downloadable NYC Local Law 144 bias audit report (dated June 12, 2025, 5,760 samples) by independent auditor Warden AI, with full sex, race/ethnicity, and intersectional disparate-impact tables (impact ratios 0.83-1.00).
- Audit reportCaptured Jun 7, 2026
Live Warden AI assurance dashboard showing monthly continuous bias auditing of Autopilot across 15 protected categories (all 'Clear'), latest audit Apr 13, 2026 with a 35,388-profile sample, covering NYC LL144, Colorado SB 205, EU AI Act 2024/1689, and California FEHA.
- Public statementCaptured Jun 7, 2026
Warden AI case study confirms continuous, independent third-party bias auditing of Juicebox's sourcing platform with a live AI Assurance Dashboard and CEO statements on responsible AI.
FRIA Support
2 items35
- AbsenceCaptured Jun 7, 2026
No FRIA template, Article 27 deployer guidance, or fundamental-rights impact assessment material was found on the trust center, privacy policy, docs site, or Warden dashboard; the EU AI Act presence is limited to a bias-audit card.
- DocumentationCaptured Jun 7, 2026
An EU AI Act (Regulation 2024/1689) bias-audit card and live assurance dashboard provide partial deployer-facing assurance material but stop short of FRIA or Article 27 deployer guidance.
Data Governance Disclosure
2 items56
- DocumentationCaptured Jun 7, 2026
Privacy policy states data is collected only from public professional sources, sensitive data (health, political affiliation) is excluded, third-party AI providers (e.g., OpenAI) may not train on personal data and must delete it within 30 days, and candidates can opt out via the Privacy Center.
- DocumentationCaptured Jun 7, 2026
Wolfia-powered trust center advertises compliance certifications, security policies, and a subprocessor list, but the underlying SOC 2 / ISO 27001 artifacts are gated behind an access request rather than publicly downloadable.
Human Oversight Design
2 items50
- DocumentationCaptured Jun 7, 2026
Audit report describes Autopilot as an assistant that scores and ranks profiles against user-set job criteria so recruiters can 'quickly focus on the best-fit talent,' indicating a human-in-the-loop decision-support model rather than autonomous decisioning.
- AbsenceCaptured Jun 7, 2026
No documentation of override controls, in-product per-candidate score explainability, per-jurisdiction compliance toggles, or audit logs was found on the product pages, docs site, or trust center.
Post-Market Monitoring
2 items55
- DocumentationCaptured Jun 7, 2026
Public Atlassian status page monitoring the Juicebox Web App and Search Service with 90-day uptime and a public incident history (no incidents reported).
- Audit reportCaptured Jun 7, 2026
Warden AI live dashboard provides continuous monthly post-market fairness monitoring of the deployed Autopilot system, with the latest audit dated Apr 13, 2026.
Customer Documentation
3 items60
- DocumentationCaptured Jun 7, 2026
Detailed public product pages for PeopleGPT and AI agents describing search and sourcing functionality, plus a docs site at docs.juicebox.work.
- DocumentationCaptured Jun 7, 2026
Services Agreement that incorporates a Data Processing Agreement by reference, supported by a Privacy Center for candidate data rights.
- AbsenceCaptured Jun 7, 2026
No dedicated customer-facing NYC Local Law 144 or EU AI Act deployer compliance guide was found, beyond the privacy policy and the Warden bias-audit dashboard.
Editorial notes
Company overview
Juicebox, operating the PeopleGPT product, is a San Francisco-based AI recruiting startup founded in 2022 by David Paffenholz and Ishan Gupta (Y Combinator S22). Its core product is a natural-language candidate search engine that indexes 800M+ public professional profiles aggregated from 30+ sources, plus an 'Autopilot' agent that automatically finds, ranks, and emails candidates against job-specific criteria. The company raised a $30M Series A led by Sequoia (≈$36M total), reports $10M+ ARR and 2,500–4,000+ customers, and serves both startups and enterprises. Pricing is roughly $139–$199 per seat/month, placing it in the mid tier.
Regulatory exposure
As an AI tool that sources and scores/ranks candidates, Autopilot is squarely an automated employment decision tool under NYC Local Law 144 and a high-risk employment AI system under the EU AI Act, Colorado SB 205, and California FEHA. Juicebox is unusually proactive here: it engages independent UK auditor Warden AI for continuous (monthly) bias auditing of Autopilot and publishes a live AI Assurance Dashboard plus a downloadable NYC LL 144 disparate-impact audit report. The main residual exposure is that no EU AI Act Article 27 FRIA materials or explicit deployer-obligation guidance are published, in-product human-oversight/explainability controls are not documented, and Article 11-style technical model documentation (model cards, ISO 42001) is absent.
Path to a higher score
Juicebox already leads its category on bias-audit transparency; to raise its overall score it should publish Article 11-style technical documentation (a model/system card and explainability statement beyond the Warden methodology), pursue ISO/IEC 42001 to anchor an AI management system, and add EU AI Act deployer guidance and a Fundamental Rights Impact Assessment template for customers. Documenting concrete human-oversight and override controls, surfacing per-candidate score explanations in-product, and making the SOC 2 / ISO 27001 artifacts referenced in its Wolfia trust center publicly verifiable would close most remaining gaps.
Regulatory frame
What applies to candidate sourcing AI.
Ranking outputs are inside Annex III §4 when they influence hiring decisions. GDPR Article 22 (automated decision-making) and the GDPR rights around profiling apply independently for EU candidates.
Compare
Juicebox (PeopleGPT) against its nearest-scoring peers.
The closest scores in Candidate sourcing AI, side by side.
- Side-by-sideJuicebox (PeopleGPT) vs Sense58F
- Side-by-sideJuicebox (PeopleGPT) vs Tezi58F
- Side-by-sideJuicebox (PeopleGPT) vs Pinpoint57F
- Side-by-sideJuicebox (PeopleGPT) vs LinkedIn Recruiter57F
Ranking
Others rated in Candidate sourcing AI
Ranked 11 of 86 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.
- #8Findem58F
- #9Pinpoint57F
- #10LinkedIn Recruiter57F
- #11Juicebox (PeopleGPT)This profile57F
- #12Rippling56F
- #13Ashby56F
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.