Tengai
Tengai runs structured screening interviews through a browser-based AI avatar and combines candidates' yes/no qualification answers, a Five-Factor-based personality score and recruiter ratings of open answers into an automatically ranked shortlist.
HireAIScore rates Tengai 42 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed September 25, 2026. Tengai ranks 110 of 160 vendors rated in Screening AI, against a category average of 46. That score is drawn from 42 evidence items across 7 rubric criteria for Tengai — 37 with a cited source, 5 recording that nothing was located.
§ 01 - Company facts
- Legal name
- Tengai AB
- Founded
- 2019
- Headquarters
- Sweden · SE
- Pricing tier
- Mid
- Market side
- Employer-side (AEDT)
- Categories
- ScreeningVideo interviewAssessment
- Website
- tengai.io
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%45
+9.0 · category avg 49
- 02
Bias Audit Transparency
Weight 18%33
+5.9 · category avg 41
- 03
FRIA Support
Weight 15%28
+4.2 · category avg 31
- 04
Data Governance Disclosure
Weight 15%50
+7.5 · category avg 53
- 05
Human Oversight Design
Weight 12%58
+7.0 · category avg 56
- 06
Post-Market Monitoring
Weight 12%33
+4.0 · category avg 40
- 07
Customer Documentation
Weight 8%55
+4.4 · 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
Human Oversight Design
Raw score 58 · contributes 7.0 to total.
58 against a 56 category average
Weakest category
FRIA Support
Raw score 28 · contributes 4.2 to total.
28 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
- 42
- Documentation
- 23
- Public statement
- 14
- Absence
- 5
- With a source URL
- 37 of 42
- Source hostnames
- 5
- Fewest items
- 5
- FRIA Support
These figures measure how thoroughly Tengai was reviewed, not how Tengai performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
8 items45
- Public statementCaptured Sep 25, 2026
Claims 'full compliance with the EU's new AI Act, where we meet all requirements' and that no AI in Tengai handles interview responses, makes assessments or takes decisions, with no supporting technical documentation.
- Public statementCaptured Sep 25, 2026
October 2025 CPO Q&A sets out the compliance rationale: AI only powers the avatar's animation, voice and language understanding, answers are scored by fixed rules, and this is said to make Tengai 'fullt kompatibel' (fully compatible) with the AI Act.
- DocumentationCaptured Sep 25, 2026
Publishes six ethical-AI principles and bias-mitigation strategies (no face recognition, NLU tuned by demographic, answer length and language quality ignored) and promises 'continuous internal tests', but gives no technical detail or results.
- DocumentationCaptured Sep 25, 2026
Explains the score components (automatic qualification and personality scores, a manual behaviour score and a 1-5 Potential Index) on a Big Five basis, but is out of step with the 4.0 scoring model in the help centre.
- DocumentationCaptured Sep 25, 2026
Swedish help article documents the 0-10 personality-match score across six facets, automatic recalculation when weights change, and the system-generated personality description shown to recruiters.
https://help.tengai.io/sv/articles/16860480-sa-tolkar-du-kandidatens-personlighetsresultat
- DocumentationCaptured Sep 25, 2026
The DPA describes processing as 'interpreting audio to text, scoring questions' and names Microsoft automatic speech recognition as the subprocessor converting streamed candidate voice to text, an AI step the compliance page's no-AI claim does not address.
- Public statementCaptured Sep 25, 2026
The candidate privacy policy says Tengai 'will conduct an AI-driven interview' and on that basis 'determine your compatibility' with the job, contradicting the compliance page's account.
- AbsenceCaptured Sep 25, 2026
No model or system card, Annex IV-style technical file, Article 13 instructions for use, Article 6 classification analysis, ISO/IEC 42001 certificate, or description of the speech-recognition/NLU models or the 4.0 question-suggestion and auto-translation features was found; searched the tengai.io sitemap (incl. /compliance, /ai-ethics, /assessment, /resources), the English, Swedish and Norwegian help centres, the DPA and all eight English blog listing pages.
Bias Audit Transparency
6 items33
- DocumentationCaptured Sep 25, 2026
The only study offered is a Psychometrics Sweden AB validation report behind a 'Free download' lead form; the page promises to show 'How unbiased Tengai's interview methodology and scoring is' but publishes no results.
- Public statementCaptured Sep 25, 2026
TNG identifies the report as a convergent and discriminant validation of extraversion and conscientiousness, commissioned by Tengai AB and begun in summer 2019 on the physical robot, and flags its own page as no longer current.
https://www.tng.se/fordomsfri-rekrytering/valideringsstudie-robotintervjuer/
- Public statementCaptured Sep 25, 2026
March 2020 press release announcing the study reports no results by gender, age or ethnicity; its 420-interview figure relates to a candidate-experience survey, not the validation sample.
- Public statementCaptured Sep 25, 2026
August 2023 Tengai press release promotes an Auburn University study of chatbot-inferred personality scores without stating that the study examined Tengai's product.
- Public statementCaptured Sep 25, 2026
Says Tengai will 'continuously do internal tests' to ensure objective outcomes, but publishes no test design, dates or results.
- AbsenceCaptured Sep 25, 2026
No NYC LL 144 bias-audit summary, independent fairness audit (BABL AI, DCI, ORCAA, Warden AI, Holistic AI, Credo AI, ConductorAI) or adverse-impact data found; searched the tengai.io sitemap, /resources hub, all blog listings, the Mynewsdesk newsroom, trust.warden-ai.com/tengai (404) and web searches for Tengai bias audits.
FRIA Support
5 items28
- DocumentationCaptured Sep 25, 2026
Commits Tengai to assist controllers with GDPR Articles 32-36 obligations, which include DPIAs, the only impact-assessment support found.
- Public statementCaptured Sep 25, 2026
June 2026 CTO Q&A gives buyers a six-question vendor checklist and notes employers' legal responsibility, but offers no FRIA template or Article 26/27 walkthrough.
https://tengai.io/blog/how-to-choose-eu-ai-act-compliant-hiring-software
- Public statementCaptured Sep 25, 2026
June 2026 post lists 'AI systems automatically evaluating or ranking candidates' as not allowed under the AI Act, a misstatement (such systems are high-risk, not prohibited) that undercuts it as deployer guidance.
https://tengai.io/blog/rectech-ready-for-the-eus-new-ai-regulations
- DocumentationCaptured Sep 25, 2026
Tengai's public-sector recruitment page, which quotes Nacka Municipality, contains no mention of the EU AI Act, fundamental rights impact assessments or DPIAs.
- AbsenceCaptured Sep 25, 2026
No Article 27 FRIA template, deployer impact-assessment worksheet or Article 26 guidance (worker-representative notice, log retention, monitoring) found on the compliance page, resources hub, help centres or blog.
Data Governance Disclosure
6 items50
- DocumentationCaptured Sep 25, 2026
Public DPA names subprocessors (Elastx private cloud in Sweden for hosting, Microsoft speech-to-text with no storage, Sinch email within the EU/EEA), lists data categories and purposes, and keeps access logs for at least five years unless instructed otherwise.
- DocumentationCaptured Sep 25, 2026
Candidate policy sets legitimate interest as the legal basis, EU/EEA processing, a 36-month maximum retention and a DPO contact, and names Intercom as its support tool, which is absent from the DPA's subprocessor annex.
- Public statementCaptured Sep 25, 2026
States Tengai avoids historical hiring data, stores only email addresses and voice recordings after interviews, and 'is certified according to ISO/IEC 27001, 27017, 27018 and 14001 standards'.
https://tengai.io/blog/how-to-choose-eu-ai-act-compliant-hiring-software
- DocumentationCaptured Sep 25, 2026
Swedish compliance page says Tengai is ISO-certified 'via vår cloud-leverantör' (via our cloud provider) and never lets external parties such as ChatGPT access candidate data.
- DocumentationCaptured Sep 25, 2026
Hosting subprocessor Elastx states it has been ISO 27001 certified since 2015 and holds 27017, 27018 and 14001, the same four standards Tengai's blog presents as its own.
- DocumentationCaptured Sep 25, 2026
Lists exclusions: no face recognition, no access to candidates' identifiable personal data during development, and no weight on answer length or language quality.
Human Oversight Design
6 items58
- DocumentationCaptured Sep 25, 2026
The 4.0 candidate card shows a total plus qualification, assessment and personality sub-scores, an interview log in question order, requirement flags on yes/no answers, and recruiter four-point ratings of text and voice answers.
https://help.tengai.io/sv/articles/16001169-sa-fungerar-det-nya-kandidatkortet-i-tengai-4-0
- DocumentationCaptured Sep 25, 2026
Recruiters pick or customise how the six personality facets are weighted (Neutral by default for ATS-created jobs), and every candidate's result is recalculated automatically, including completed interviews.
https://help.tengai.io/sv/articles/16860425-sa-viktar-du-personlighetsprofilen-i-tengai
- DocumentationCaptured Sep 25, 2026
Candidates hear the system's personality description at the end of the interview and comment on it, and the unedited comment appears beside the system description on the recruiter's card.
https://help.tengai.io/sv/articles/16860480-sa-tolkar-du-kandidatens-personlighetsresultat
- Public statementCaptured Sep 25, 2026
The 4.0 release keeps rule-based scoring and logs every mid-process change, but tells recruiters they need not listen to every answer because 'our shortlist automatically surfaces the top performers'.
https://tengai.io/blog/tengai-4.0-structured-flexible-screening
- Public statementCaptured Sep 25, 2026
Tells candidates that a human, 'not AI, bots, or any other algorithm', always evaluates their interview, and to contact the inviting employer if a disability makes a verbal or written interview difficult.
- Public statementCaptured Sep 25, 2026
Claims psychometric experts validated that Tengai 'independently can screen and recommend job-seekers, without any human interference'.
Post-Market Monitoring
5 items33
- AbsenceCaptured Sep 25, 2026
No public status page, incident history, security.txt, vulnerability-disclosure policy or security contact: status.tengai.io and trust.tengai.io resolve to a Loopia domain-parking page (identical for random subdomains), tengai.statuspage.io redirects to Atlassian's marketing site, and tengai.io/status, /security, /trust and /.well-known/security.txt return 404.
- DocumentationCaptured Sep 25, 2026
Commits to written notice of personal-data breaches without undue delay, at least annual security self-monitoring with results available to the customer on request, and logging of access to personal data.
- DocumentationCaptured Sep 25, 2026
Dated help article (20 July 2026) records a scoring change in Tengai 4.0: the four fixed competencies were removed and recruiters now rate their own free-text and voice questions on the same four-point scale.
https://help.tengai.io/sv/articles/16001048-sa-fungerar-bedomning-i-nya-tengai
- Public statementCaptured Sep 25, 2026
Product changes are announced through release posts such as 4.0 (17 September 2026), which says every mid-process change is logged automatically; no separate model or scoring changelog exists.
https://tengai.io/blog/tengai-4.0-structured-flexible-screening
- DocumentationCaptured Sep 25, 2026
The only published channel for candidates' data concerns is privacy@tengai.io (the DPO), with complaints escalated to Sweden's IMY.
Customer Documentation
6 items55
- DocumentationCaptured Sep 25, 2026
How-it-works page covers assignment set-up, invitations, NLU-based avatar interviews and the scored shortlist, with recruiter and candidate FAQs.
- DocumentationCaptured Sep 25, 2026
Full GDPR Article 28 DPA text, with processing instructions and a subprocessor annex, is published openly rather than on request.
- DocumentationCaptured Sep 25, 2026
Swedish help centre walks customers through the candidate journey, including invitations, reminders, recruiter-chosen answer modes, a voice-only closing question and reuse of earlier personality profiles.
- DocumentationCaptured Sep 25, 2026
Candidate guide explains the avatar, its use of AI speech recognition, answer formats, the feedback candidates receive and next steps.
- DocumentationCaptured Sep 25, 2026
The English help centre states it has no articles or collections yet, so product documentation exists only in Swedish and partly in Norwegian.
- AbsenceCaptured Sep 25, 2026
No Article 13-style instructions for use, EU AI Act deployer guide, or NYC LL 144 or other US-jurisdiction guidance found on the site, the resources hub or the help centres.
§ 05 - Editorial notes
Company overview
Tengai AB is a Stockholm company that grew out of staffing firm TNG's innovation lab in 2017, built its first interviewer on Furhat Robotics' social-robot platform in 2018 and was incorporated as a separate company in 2019; in 2022 the physical robot was replaced by a browser-based avatar. Candidates complete a structured interview of about 20 minutes, answering by buttons, text or voice in Swedish, English, Norwegian or Danish. Recruiters get an automatically ranked shortlist built from an automatic qualification score, a 0-10 personality-match score across six facets grounded in the Five-Factor Model, and their own four-point ratings of open answers. Tengai 4.0 (September 2026) lets recruiters write their own question sets and weightings. Named customers are Nordic employers and public bodies such as Byggmax, SEB, Region Gotland and several Swedish municipalities, and it integrates with ATSs including Teamtailor, Varbi and Jobylon. Swedish registry data show a small company (six employees and about SEK 1 million revenue in 2024), and no pricing is published.
Regulatory exposure
Software that interviews, scores and ranks job candidates is the core recruitment use case in EU AI Act Annex III point 4, so as an EU provider Tengai faces the Article 9-15 requirements, conformity assessment and EU-database registration from 2 December 2027 under the Digital Omnibus timetable. Its 'Fully EU AI Act compliant' claim rests on an argument, not documents: AI is said to drive only the avatar's animation, voice and language understanding, while answers are scored by fixed 'mechanical' rules. That is contestable - spoken answers are transcribed by Microsoft speech recognition and interpreted by Tengai's NLU, and personality scoring is profiling, which under Article 6(3) keeps an Annex III AI system high-risk - and Tengai publishes no classification analysis, technical file or instructions for use. A June 2026 blog post even tells buyers that AI ranking of candidates is 'not allowed' under the Act, which regulates rather than bans it. Its municipal and regional customers would carry Article 26 deployer duties and, as public bodies, Article 27 fundamental rights impact assessments. Tengai says it uses no facial recognition and does not judge tone of voice or dialect, so no Article 5(1)(f) emotion-recognition signal was found. No US deployment was found; a US employer would need to check NYC Local Law 144 (no bias audit exists), Illinois HB 3773, California's FEHA ADS rules (four-year record retention versus Tengai's 36-month cap) and Colorado SB 26-189 from 1 January 2027.
Path to a higher score
The biggest gain is to back the 'Fully EU AI Act compliant' slogan with documents: publish an Article 6 classification analysis explaining why speech recognition, NLU and the 4.0 question-suggestion and translation features sit outside candidate evaluation, or accept high-risk status and publish an Annex IV-aligned technical summary and Article 13 instructions for use well before December 2027. Tengai should also reconcile its conflicting accounts of where AI is used (compliance page, AI-ethics page, candidate privacy policy, candidate guide) and correct the blog's claim that AI ranking is banned. On fairness, it should ungate the 2020 Psychometrics Sweden report and commission an independent adverse-impact audit of the current avatar and 4.0 scoring, broken down by gender, age, ethnicity and disability and by voice versus text answers, with published selection-rate tables. Smaller fixes: say plainly that the ISO certificates belong to hosting provider Elastx (or obtain its own ISO 27001 and ISO 42001), update the DPA's 2010 model clauses and subprocessor list, give public-sector customers FRIA and Article 26 guidance, fill the empty English help centre, and add a status page, a security contact and a scoring changelog.
§ 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→Tengai against its nearest-scoring peers.
In Screening AI.
§ Others rated in Screening AI
All screening vendors→Ranked 110 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.
- 107Talogy42F
- 108Humanly42F
- 109Recruitee42F
- 110TengaiThis profile42F
- 111hireEZ42F
- 112softgarden42F
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.