Alex

Alex handles early recruiting steps: resume screening, candidate outreach, interview coordination, conversational interviews, reports, ATS updates, and fraud signals. Evaluate one complete workflow—from ATS stage and candidate notice through interview evidence and recruiter decision—because the handoffs matter as much as the interview itself. [S1][S2]

Candidate records, interview content, scores, and fraud signals can affect employment decisions. Alex says it does not automate hiring decisions, automatically reject candidates without human input, or prevent a hiring team from overriding its output. Buyers should reproduce those controls in their own ATS workflow; an integration or score does not make the process compliant. [S3]

Best for: Talent-acquisition teams and staffing firms with high applicant volume, an owned ATS process, structured role criteria, and recruiters who retain decision authority.

Not for: teams seeking an automatic accept/reject engine, teams without a clear candidate-notice, consent, and recruiter-review process, or low-volume specialized searches where a human-led first conversation provides the necessary signal and care.

Run a first-round candidate interview

• Trigger: A candidate enters the buyer's configured interview stage or is invited from an enabled workflow.
• Inputs: The job's interview design and criteria, candidate contact details, the candidate's responses, and the buyer-permitted ATS context.
• Output: An interview record, evaluation output, and a next-step queue for the recruiting team.
• Human checkpoint: A recruiter or hiring manager reviews the record and decides whether to advance, hold, request another interview, or reject.
• Success measures: invitation-to-completion rate, time to completed first screen, candidate drop-off, recruiter agreement or override rate, candidate feedback, and adverse-impact monitoring where lawful and appropriate.

Alex says it conducts personalized candidate interviews as an always-on first-round interviewer and supports phone, video, text, and WhatsApp channels. Its ethics statement says final hiring decisions remain with the hiring team. [S1][S3]

Return a completed interview to Greenhouse

• Trigger: A recruiter moves a candidate into the Alex stage in a connected Greenhouse job and sends the invitation.
• Inputs: Greenhouse job and candidate records, the selected Alex interview configuration, the integration API key, and the candidate's completed interview.
• Output: A completed Alex stage with the candidate's score and detailed report uploaded to Greenhouse.
• Human checkpoint: The recruiter or assigned grader inspects the report and decides the next workflow action; any missing, duplicate, or incorrect record goes to the integration owner.
• Success measures: successful invitation delivery, completion and sync rate, report accuracy, duplicate-record rate, time to recruiter review, manual reconciliation, and candidate correction requests.

Alex's Greenhouse documentation describes this stage-based flow and says its setup requires an API key to be handled by the customer's Greenhouse account manager; the vendor says configuration can take 24–48 hours. [S2]

Screen and prioritize inbound applicants

• Trigger: A new application or a batch of resumes arrives for a role.
• Inputs: Resume content, the buyer's job requirements, screening criteria, and any ATS data the buyer elects to connect.
• Output: A prioritized group for human review or a configured interview workflow, with evidence the reviewer can inspect.
• Human checkpoint: The role owner checks the defined criteria, score distribution, and sampled outcomes before candidates are progressed or rejected.
• Success measures: screening turnaround, recruiter agreement, false-negative and false-positive samples, review coverage, time-to-interview, candidate experience, and disparate-impact signals.

Alex presents resume screening as a product that evaluates candidates against role-specific criteria. That is a vendor capability statement; buyers should validate its evaluation evidence and non-selection controls against the exact role and jurisdiction. [S1][S3]

The expected operating model is candidate or ATS event → configured invitation or screening → Alex interview or evaluation → score, report, and/or workflow update → recruiter or hiring-manager decision. The Greenhouse guide specifies that an Alex interview can become a stage in the buyer's pipeline and that completion returns an overall score, evaluation summary, skill scores, skill summaries, and a report link. [S2]

Alex states that it provides documented interview records and supports disclosure and consent flows. It also states that teams can challenge, override, or disregard outputs and that no candidate is automatically rejected by AI without human input. These are important assertions, but the buyer should reproduce them through configuration review, test candidates, role-specific sample cases, and a logged decision trail before relying on them. [S3]

Verified facts from primary sources

• Alex's public site describes functions for resume screening, AI interviews, candidate verification, interview coordination, ATS talent matching, and candidate interaction across phone, video, text, and WhatsApp. [S1]
• Alex documents a Greenhouse workflow where a completed candidate interview returns a score and detailed report to Greenhouse. [S2]
• The privacy policy identifies Apriora Inc. dba Alex and describes potential processing of contact, employment, resume, experience, background, education, device, and usage information depending on the use of the service. [S4]
• Public pricing was not displayed on the reviewed product, documentation, privacy, ethics, or security pages as of this research date. Request the unit, annual minimum, implementation fees, volume treatment, overages, renewal terms, and termination assistance directly from Alex.

Vendor claims to validate

• Alex states that it does not automate hiring decisions, does not automatically reject candidates without human input, and permits output overrides. [S3]
• Alex states that it is SOC 2 Type II and GDPR compliant and that its platform has encryption in transit and at rest, access controls, logging, and monitoring. Obtain the current report, certification scope, exceptions, DPA, and contractual application. [S5]
• Alex states that it runs monthly AI bias-audit checks and provides visibility into the reviews. Confirm the exact methodology, population, sampling, protected groups, version coverage, thresholds, remediation process, and whether the evidence applies to the buyer's job and configuration. [S5]

B2Bagents assessment

Alex is most credible as a recruiting-workflow system with meaningful automated execution but mandatory human decision ownership. That assessment follows from the documented ATS writeback and interviewing flow, paired with the vendor's explicit statement that hiring teams—not Alex—retain final decisions. Its value should be measured by a short, controlled pilot that tests both operational throughput and candidate-protection outcomes; a high completion or fast scheduling rate is not enough on its own. [S1][S2][S3]

Material unknowns

• Current commercial model, minimum contract, usage measure, implementation scope, support commitment, renewal treatment, and exit assistance.
• Exact data retention, deletion timing, residency options, subprocessor and model-provider list, data-training terms, buyer access to audit records, and privacy-rights handling for a particular deployment.
• Exact ATS permissions, write operations, webhook failure behavior, reconciliation method, record export, and rollback boundary for every supported integration.
• Role- and configuration-specific accuracy, score explainability, fraud-signal performance, accessibility, language coverage, candidate accommodation path, and fairness or disparate-impact results.
• Jurisdiction-specific legal obligations for disclosure, consent, recordkeeping, employment decisions, and automated-decision tools.

Start with a high-volume, clearly structured role whose candidate flow already has stable stages and a named recruiter owner. Configure a single ATS integration and one interview stage. Before inviting real candidates, document every notice, consent, message, criterion, score, reviewer, write action, exception route, retention setting, and candidate-help path.

Establish a baseline for recruiter time, invitation delivery, time to interview, completion, candidate drop-off, candidate feedback, time to review, recruiter override rate, manual data reconciliation, and time-to-hire. Where lawful and appropriate, define an approved fairness-monitoring plan before launch. Run deliberately difficult cases: ambiguous answers, non-native accents, accommodation requests, candidate corrections, duplicate candidates, no-shows, unavailable calendars, webhook failures, mixed-language conversations, false fraud signals, and score disagreement between reviewers.

Agree success thresholds before the pilot: a sufficient completion rate and time-to-review improvement without a material decline in candidate feedback; reliable and reconcilable ATS records; reviewable evidence for every output; no automatic employment action; acceptable recruiter agreement and override behavior; and no unacceptable fairness, accessibility, privacy, or fraud-signal outcome. Stop the pilot for a missing consent or audit record, an unintended candidate disposition, unexplainable scoring shift, inaccessible workflow, unreconciled ATS write, material data concern, or any concerning legally permissible disparate-impact signal.

• Juicebox: an existing directory entry focused on talent sourcing and search; compare candidate discovery and outreach with Alex's interview and workflow-execution boundary.
• Mercor: an existing directory entry that matches and evaluates talent; compare marketplace or matching-model considerations with a buyer-operated ATS workflow.
• Jack & Jill: an existing directory entry in talent operations; compare managed or service-oriented recruiting support with Alex's software platform and integration model.

1. Which actions can Alex take automatically in our ATS, calendars, and messaging tools, and which actions are always limited to human approval?
2. What precise evidence reaches the reviewer after an interview, how can it be challenged or corrected, and how do we export it for audits or candidate requests?
3. How are candidate disclosure, consent, withdrawal, accessibility, language, and accommodation handled in every enabled channel and jurisdiction?
4. What data does each integration read, create, retain, share, or use for models; who can access it; and how do deletion and export work at contract end?
5. What are the role-specific score, fraud-signal, override, and fairness results under our configuration—and what happens when a result is uncertain or contested?
6. What does the contract charge for implementation, recruiter seats, interviews, channels, integrations, verification, support, usage, overages, and renewal?