Apriora

What is Apriora, now Alex?

Apriora is the former name of the recruiting product now marketed as Alex. Alex’s privacy policy identifies the business as Apriora Inc. doing business as Alex, while the current product site is alex.com. The product is not positioned as a simple one-way video questionnaire: Alex says it conducts structured, role-specific first-round conversations, asks follow-up questions in real time, and provides the hiring team with a recording, transcript, and notes. The company says that final hiring decisions remain with people. Source: Alex Candidate FAQ; brand relationship: Privacy Policy, checked 20 August 2026.

Buyers should assess the wider workflow, not only the interview. Alex presents its platform as covering outreach, resume screening, interviewing, identity and fraud signals, scheduling, and rediscovery of prior candidates in an ATS. Conversational screening is therefore central, but it sits inside a broader automation product. An organisation seeking only a lightweight interview add-on should ask whether the desired channels and modules can be bought separately; a public packaging matrix is not available. Source: Alex platform page, checked 20 August 2026.

What does the product do technically?

Alex’s documentation says an interview can start from a job description, a template, or a manual build. Some ATS integrations can supply the job through a job ID. The system generates a draft from the job description and selected template; a recruiter can then edit, reorder, remove, or add questions. In practical terms, the automation combines a configured interview structure, role context, and evaluation instructions rather than simply playing back a fixed video script. Source: Alex documentation, Creating an AI Interview, checked 20 August 2026.

• Conversational data capture: Alex describes live interviews by video, phone, SMS, and WhatsApp. It says video and phone conversations can be recorded, transcribed, and scored, while SMS and WhatsApp support asynchronous text or media interactions. Candidates can participate outside normal business hours. Source: Alex platform page, checked 20 August 2026.
• Evaluation against defined criteria: Teams can assign criteria, priorities, and instructions to each question, including what should count as a strong or weak answer. Alex can suggest criteria from the job description, but users can change weights and instructions. A score is therefore a product of the configured questions and rubric, not a free-standing judgement of a person. Source: product documentation, checked 20 August 2026.
• Informational tags: Tags can capture operational facts such as work authorisation, location, availability, or experience. The documentation says they do not automatically affect the overall score. This separates reportable facts from evaluation, provided that the interview actually asks a question capable of surfacing the information. Source: product documentation, checked 20 August 2026.
• Recruiting data sources: Alex describes screening resumes against role criteria, using application materials and interview content, and searching prior candidates already held in an ATS through Talent Match. Its candidate FAQ says Alex does not scrape candidate data from sites such as LinkedIn or GitHub and does not build external profiles from third-party datasets. Source: Alex Candidate FAQ, checked 20 August 2026.
• Workflow automation: Once a candidate clears a configured bar, Alex says its Coordinator can book the next interview directly on team calendars, prepare a briefing with scores and insights, and handle confirmations, reminders, and rescheduling. The intended workflow is thus application through to a human handoff. Source: Alex platform page, checked 20 August 2026.
• Identity and fraud signals: The Verify module is described as analysing audio, video, and behavioural signals to flag external assistance, impersonation, and synthetic profiles. This is the vendor’s description of product functionality, not an independent validation of accuracy, fairness, or suitability for a particular role or population. Source: Alex platform page, checked 20 August 2026.

Automation does not remove process design

The potential automation depth is substantial: applications can be screened immediately, candidates can be contacted or invited, interviews can run around the clock, results can feed back into the workflow, and qualified people can be scheduled. Yet output quality still depends on four human choices: the job definition, the questions, the evaluation instructions, and the threshold for escalation. Alex’s documentation explicitly notes that it can only score a criterion where the interview includes a question that elicits the relevant information. Source: product documentation, checked 20 August 2026.

Candidate context matters as well. The configuration allows a team to supply role and company information for Alex to use when candidates ask questions at the end of an interview. If the field is empty, Alex intentionally does not ask whether the candidate has questions, to avoid exposing sensitive information. A buyer therefore needs an explicit content decision: what may Alex say about the role, team, working arrangement, process, and employer? Source: product documentation, checked 20 August 2026.

The structural limitation is not only model performance. A standardised conversation can make evidence easier to compare, but it will not automatically create rapport, answer every nuanced culture question, or interpret an unusual career story well. For scarce technical roles, senior hires, or candidate groups for whom a conversational AI is an unfamiliar first contact, a pilot should test perception and define when a human recruiter takes over earlier.

When is Apriora / Alex a good choice?

Apriora / Alex is a good choice for teams with many repeatable first-round conversations, clear baseline criteria, and an ATS that should receive workflow outcomes. That can include staffing and RPO organisations, high-volume role families, and internal TA teams that want recruiters to spend less time on scheduling and initial screens and more time on higher-value conversations. The value is not merely an AI conversation; it comes from combining a consistent interview guide, documented evidence, and an automated handoff. Source: Alex platform page, checked 20 August 2026.

Pricing, contract scope, and economics

Alex does not publish prices. The official website directs buyers to a demo request, and the G2 listing shows “Contact Us.” The reviewed public material does not state package prices, consumption-based interview prices, implementation charges, or a minimum term. It would therefore be unsound to infer an annual licence fee or a per-interview rate. Source: Alex demo page; G2 pricing listing, checked 20 August 2026.

An offer should distinguish more than a single total. Ask separately about interview channels, interview and application volume, Resume Screens, Verify, Talent Match, ATS connectors, calendar and communications connections, implementation, configuration, migration, training, support tier, and contract length. The documentation notes that some accounts only have access to one interview modality according to their subscription, so the availability of each channel is a concrete commercial question. Source: product documentation, checked 20 August 2026.

Rather than assuming that fees are hidden, require the proposal to state every potential cost element: one-time implementation, workflow and rubric configuration, additional ATS or communications connectors, data retention, overage volume, security review, SSO, language testing, change management, and candidate support. The same conversation should cover notice periods, minimum commitments, treatment of unused volume, and the price of later expansion to more countries or business units.

A decision rule that does not depend on guessed team size

Alex does not publish the team size at which it pays back. A more defensible rule is to calculate: break-even applications = annual total cost / (recruiter minutes genuinely removed per application × fully loaded recruiter hourly cost / 60). Annual total cost should include licence, implementation, administration, and support. Removed minutes should come from a buyer’s own pilot and only include work that actually disappears, such as scheduling, first screens, documentation, and follow-up.

This means a smaller team with several hundred similar applications can have a clearer business case than a larger company making only a few highly bespoke executive hires. The pilot should measure candidate completion, technical failures, human rework, time to handoff, and acceptance by channel. Those operational numbers show whether automation releases recruiter capacity or simply moves work into configuration and exception handling.

What do users and public sources report?

When checked, G2 displayed 29 reviews and a 4.8/5 score. Visible reviewers from recruiting and staffing praised handling larger applicant pools, ATS integration, and the opportunity to capture information beyond the resume. Those observations are consistent with the vendor’s high-volume workflow position, but they are not an independent effectiveness study. Several visible reviews are labelled as collected through a seller invitation, so the small sample, collection method, and relevant role should be considered when interpreting the score. Source: G2 reviews, checked 20 August 2026.

A contrasting public account comes from a 2025 Futurism report. It describes one candidate’s and one tester’s experience of generic questions, slow responses, and missing information about the purported role. It does not establish a general failure rate or represent every deployment. It does support a practical pilot requirement: invitations must be clear, the role context sufficient, support reachable, and a human fallback available when an automated live conversation goes wrong. Source: Futurism, published 17 September 2025, accessed 20 August 2026.

Alex also publishes customer and candidate testimonials and scale or preference figures on its own website. These are useful evidence of the intended use case, but they are vendor material. A buyer should supplement them with reference calls for comparable role types and a candidate pilot in its own process. Source: Alex home page, checked 20 August 2026.

DACH, EU, GDPR, and the EU AI Act

For DACH roles, language coverage is only a starting point. Alex’s home page says 30 supported languages and its platform page says 26+. The reviewed material does not show which release, dialects, role vocabulary, or accessibility features are included in a particular contract. Alex presents candidate testimonials concerning accents, but no independent German-language quality measurement, DACH support commitment, or regional reference data was found in the reviewed sources. German-language roles should be tested with realistic accents, terminology, and candidate populations. Source: Alex home page; platform page, checked 20 August 2026.

Alex states on its platform page that it has SOC 2 Type II, GDPR compliance, human oversight, and monthly bias audits by Warden AI across more than 15 protected classes. Its privacy policy sets out GDPR legal bases, EEA data-subject rights, and an EU representative in Bonn. The same policy says personal data may be transferred to countries other than where it was collected and does not identify a specific hosting location. That is not evidence against EU hosting, but neither is it a public EU-hosting commitment. The processor agreement, subprocessors, data region, transfer mechanism, and deletion periods should be assessed contractually. Source: Alex platform page; privacy policy, last updated 6 February 2025, accessed 20 August 2026.

The EU AI Act assessment depends on the exact intended use. The European Commission identifies AI used for employment, worker management, and access to self-employment – including CV-sorting software for recruitment – as a high-risk area. From 2 December 2027, high-risk systems in those areas are subject to strict requirements including risk management, logging, technical documentation, information for deployers, human oversight, and requirements for robustness and accuracy. Alex says its bias audits align with the EU AI Act and that it does not automatically reject candidates. That does not replace diligence on whether the specific configuration, scoring weights, and employer deployment meet the applicable obligations. Source: European Commission, EU AI Act; Alex platform page, checked 20 August 2026.

Eight questions for the selection conversation

1. Which components and channels are included in the proposed plan, and what minimum commitment, volume commitment, or contract term applies to each?
2. In which region are resumes, audio, video, transcripts, scores, and Verify signals processed and stored; which subprocessors and transfer mechanisms apply?
3. Which of the 33 ATS integrations is included in the contract, which fields sync in each direction, and how are failed syncs, duplicates, and interrupted interviews handled?
4. How are job descriptions, questions, follow-up logic, weights, and decision thresholds reviewed, approved, and versioned before they go live?
5. Which Verify functions are enabled, what signals do recruiters see, what false-positive evidence exists in the relevant context, and what human review is mandatory before any consequence?
6. How will German performance be tested for the planned roles, including vocabulary, accents, captions, assistive technology, and phone, video, SMS, or WhatsApp?
7. What audit evidence, bias-test methodology, group-level performance information, and documentation for DPIA or EU AI Act work can Alex provide?
8. What candidate alternative and support process applies when a person declines an AI interview, a link fails, or the conversation quality is inadequate?

Frequently asked questions

Does Alex automatically reject candidates?

Alex says no: it shares recordings, transcripts, and notes while humans make hiring decisions. A pilot should nevertheless test which workflows a low score triggers and whether recruiters can meaningfully override or disregard outputs. Source: Alex Candidate FAQ, checked 20 August 2026.

Is it just a video questionnaire?

No. Alex describes a responsive conversation that can ask follow-up questions. The interaction remains bounded by the configured questions, scoring criteria, and employer-provided context, which is why configuration quality matters. Source: Alex Candidate FAQ; product documentation, checked 20 August 2026.

Can a candidate recover from a technical problem?

The candidate FAQ describes support and the option to use the same interview link at another suitable time. Buyers should also agree a tested escalation route, response times, and a human alternative. Source: Alex Candidate FAQ, checked 20 August 2026.

Are German support and EU hosting guaranteed?

Publicly evidenced are a multilingual product position, GDPR-related statements, and an EU representative. A specific German-quality commitment and a named EU hosting location are not publicly evidenced, so they belong in contractual documentation, security materials, and pilot acceptance criteria. Source: Alex platform page; privacy policy, checked 20 August 2026.

Which candidate data does Alex use?

Alex describes using interview content and application materials, while its ATS features use existing candidate records. The candidate FAQ says it does not scrape third-party sites such as LinkedIn or GitHub or create external profiles from them. Source: Alex Candidate FAQ, checked 20 August 2026.

How long should a pilot run?

No public pilot duration is specified. It should include enough real candidates and at least one complete ATS workflow to measure acceptance, completion, exceptions, rework, and time saved against the current process. The decision should follow those measurements and the actual proposal, not a generic headcount rule.

Overall, Apriora / Alex is most compelling where conversational first screens sit inside a defined, ATS-connected workflow with meaningful human oversight. For DACH or wider EU deployments, data region, contract evidence, language performance, candidate alternatives, and governance are core acceptance criteria rather than secondary legal checks.