Greenhouse

AI features

As the vendor or firm describes them. Notes in small print record what the sources leave open.

1. Talent Matching

   Analyzes each resume against recruiter-defined, weighted job criteria, sorts applications into tiers and explains which criteria a candidate matches; it does not advance or reject candidates.

   Kind

   Built into the platform

   Suggests or acts

   Suggests

   Who controls it

   Customer HR team

   Availability

   Generally available

   Control: Recruiters set and weight the criteria; customers can switch the feature on or off at the organization level and by office, department or job.

   4 sourcesSourcesTalent Matching, bias audits and transparency (bias audit statement)Real Talent (talent matching, fraud detection and identity verification product page)Our AI principlesGreenhouse Plans & Pricing | Core, Plus & Pro

Unclear or conflicting statements

Where the sources are unclear or disagree, both statements are shown with their links.

Questions to ask

For this offering

• Which version of Talent Matching runs in our account, when did it last pass Warden AI's audit, and will we be told when the model or its large language model provider changes?
• The published audit uses Warden AI's own test profiles: what adverse-impact data can we export from our own applicant flow, by job and stage, to run our own analysis?
• How are the tier categories derived from the criteria weights, and is there a numeric match score underneath them that recruiters or reports can see?
• Which settings (criteria, weights, enabled offices or jobs) change who appears in the top tiers, who in our organization can change them, and is each change logged?
• How does a candidate's request for manual review reach a recruiter, and what record is kept of the review?
• How long are the data underlying match scores retained, and can retention be set to our legal record-keeping period?

From the diligence question set

1. DQ-37 Has an independent auditor tested this tool for adverse impact, on data like ours? Can we see the full results: the groups tested and excluded, the data and its dates, the thresholds used, and the auditor's relationship to you?
2. DQ-38 Can we monitor adverse impact on our own applicant flow, by job and by stage, and export the data to do it?
3. DQ-39 What evidence shows that the tool measures something related to the job, and was it validated for the way we will use it: as a pass-or-fail screen, or to rank people?
4. DQ-40 Which settings change who passes: weights, thresholds, knockout questions, cut scores? Who can change them, are changes logged, and does a change trigger a new audit?
5. DQ-41 What notices, explanations and alternative processes does the tool support, and can a person with authority review and change an adverse decision?
6. DQ-44 What will you commit to in the contract to support our obligations: audit cooperation, documentation of intended use and limits, notice support, records, and cooperation if a decision is challenged?

Build questions for this vendor

AI features

As the vendor or firm describes them. Notes in small print record what the sources leave open.

1. Voice AI interview

   Runs a structured two-way voice interview with dynamic follow-up questions and answers to candidate questions drawn only from information the hiring team approved.

   Kind

   Agents and automated actions

   Suggests or acts

   Acts

   Who controls it

   Customer HR team

   Availability

   Generally available

   Control: Recruiters and hiring managers set the questions, rubric and role context.

   1 sourceSourceAI interview agent (Greenhouse Voice AI product page)

2. Interview scoring and explanations

   Scores each answer against the customer's rubric and examples, with detailed explanations; scores are based on audio transcripts and can be overridden.

   Kind

   Agents and automated actions

   Suggests or acts

   Suggests

   Who controls it

   Customer HR team

   Availability

   Generally available

   Control: Managers can calibrate scores; recruiters and hiring managers can override them.

   1 sourceSourceAI interview agent (Greenhouse Voice AI product page)

3. Cheat and inconsistency detection

   Flags inconsistencies, AI-generated answers and signs that a candidate is getting outside help, timestamped in the transcript for the team to review.

   Kind

   Agents and automated actions

   Suggests or acts

   Suggests

   Who controls it

   Not stated

   Availability

   Generally available

   Control: Not stated.

   1 sourceSourceAI interview agent (Greenhouse Voice AI product page)

Unclear or conflicting statements

Where the sources are unclear or disagree, both statements are shown with their links.

Questions to ask

For this offering

• Can we see the latest Warden AI results for Voice AI, the data they used and the protected classes covered, and how often the scoring model changes between audits?
• Is Voice AI inside the scope of Greenhouse's current ISO/IEC 42001 certificate today, and can we see the certificate and its scope statement?
• Which large language model provider processes the interview audio and transcripts, and how long are recordings and transcripts kept?
• How does a candidate who cannot or prefers not to complete a voice interview request an alternative or an accommodation, and does speech or response timing affect any flag or score?
• What do the cheat and AI-answer flags rely on, how often are they wrong, and are they shown to hiring managers before or after the score?
• What notice text do candidates see before the interview starts, and can we edit it for each jurisdiction?

From the diligence question set

1. DQ-37 Has an independent auditor tested this tool for adverse impact, on data like ours? Can we see the full results: the groups tested and excluded, the data and its dates, the thresholds used, and the auditor's relationship to you?
2. DQ-38 Can we monitor adverse impact on our own applicant flow, by job and by stage, and export the data to do it?
3. DQ-39 What evidence shows that the tool measures something related to the job, and was it validated for the way we will use it: as a pass-or-fail screen, or to rank people?
4. DQ-41 What notices, explanations and alternative processes does the tool support, and can a person with authority review and change an adverse decision?
5. DQ-42 How can a candidate with a disability take the assessment in an accessible format or request an alternative, and does speed, voice, video or a timed task affect the score?
6. DQ-43 Does any feature infer emotion, personality or other traits from face, voice or video, or use biometric data? Can each such feature be switched off?
7. DQ-44 What will you commit to in the contract to support our obligations: audit cooperation, documentation of intended use and limits, notice support, records, and cooperation if a decision is challenged?

Build questions for this vendor