softgarden

softgarden is a German cloud applicant tracking system (multiposting, career sites, applicant management and onboarding) with optional AI features: AI-Assisted Applicant Matching, which highlights applicants who closely match a job's requirements, and a newer LLM-based Screening Assistant, which scores each applicant against screening criteria.
ATS-integratedScreening

HireAIScore rates softgarden 42 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed September 25, 2026. softgarden ranks 57 of 77 vendors rated in ATS-integrated AI, against a category average of 47. That score is drawn from 36 evidence items across 7 rubric criteria for softgarden — 31 with a cited source, 5 recording that nothing was located.

Score breakdown

softgarden, 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.

  1. Article 11 Technical Documentation

    Weight 20% · avg 51

    42

    +8.4

  2. Bias Audit Transparency

    Weight 18% · avg 41

    22

    +4.0

  3. FRIA Support

    Weight 15% · avg 32

    30

    +4.5

  4. Data Governance Disclosure

    Weight 15% · avg 54

    54

    +8.1

  5. Human Oversight Design

    Weight 12% · avg 57

    58

    +7.0

  6. Post-Market Monitoring

    Weight 12% · avg 42

    44

    +5.3

  7. Customer Documentation

    Weight 8% · avg 59

    58

    +4.6

Total weighted score

41.8(rounded to 42 for display)

Category avg is the mean raw score on that criterion across the 77 ATS-integrated AI vendors in scope of this rubric, this one included.

Strongest criterion

Customer Documentation

58raw / 100

Contributes 4.6 to the total. 58 against a 59 category average.

Weakest criterion

Bias Audit Transparency

22raw / 100

Contributes 4.0 to the total. 22 against a 41 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 record

Evidence ledger

Items
36
Documentation
24
Audit report
3
Public statement
4
Absence
5
With a source URL
31 of 36
Source hostnames
4
Fewest items
3
Customer Documentation

These figures measure how thoroughly softgarden was reviewed, not how softgarden performed — an absence row, recording that nothing was located, is counted like any other item.

Article 11 Technical Documentation

7 items42
  • DocumentationCaptured Sep 25, 2026

    The help article says the AI matching was 'initially categorized as high-risk under Annex III, No. 4' but claims exemption under Article 6(3), points (a) and (d), because it 'supports decision-making rather than replacing human evaluation'; it does not publish the underlying Art. 6(4) assessment.

    https://support.softgarden.de/en/articles/319436-what-is-the-ai-assisted-applicant-matching

  • DocumentationCaptured Sep 25, 2026

    The linked 'Technische Beschreibung Matching Solution' PDF runs to two pages covering purpose, an anonymised training dataset, coordination with the external data protection officer, opt-in activation, privacy-notice wording and UI highlighting; it contains no model design, input features, accuracy or validation results, subgroup performance or known limitations.

    https://support.softgarden.de/de/articles/319436-ki-gestutztes-bewerber-matching

  • DocumentationCaptured Sep 25, 2026

    softgarden's own applicant-profile help article presents the feature under the heading 'Easier Selection with AI Matching', which sits awkwardly with its claim that the system performs only narrow procedural or preparatory tasks.

    https://support.softgarden.de/en/articles/324356-what-does-the-applicant-profile-include

  • DocumentationCaptured Sep 25, 2026

    The 9 May 2025 launch note invited customers to let softgarden AI help 'die Top-Bewerber aus der Menge herauszufiltern' (filter the top applicants out of the crowd), wording that mirrors Annex III 4(a)'s 'analyse and filter job applications'.

    https://support.softgarden.de/de/articles/472138-produkt-updates-q2-2025

  • Public statementCaptured Sep 25, 2026

    The pricing FAQ says softgarden uses AI 'when selecting applicants (matching)' and calls its AI 'Transparent and without a black box', with no technical documentation behind that claim.

    https://softgarden.com/en/pricing/

  • Public statementCaptured Sep 25, 2026

    The page for the AI Screening Assistant, an LLM feature that gives each applicant an overall score against screening criteria, states only that it is 'GDPR-compliant and in line with the EU AI Act', with no risk classification, technical documentation or instructions for use.

    https://softgarden.com/en/products/ai-screening-assistant/

  • AbsenceCaptured Sep 25, 2026

    No model or system cards, Annex IV-style technical pack, instructions for use or ISO/IEC 42001 certification were found in the softgarden.com sitemap (661 URLs), the support.softgarden.de sitemap (675 URLs), the help center's own search for 'AI Act', 'KI-Verordnung' and 'Hochrisiko', the data-protection and certificates page, the DPA or the code of ethical conduct.

    Have a URL that disputes this?

Bias Audit Transparency

4 items22
  • AbsenceCaptured Sep 25, 2026

    No NYC LL 144 bias audit, independent algorithmic audit (BABL AI, DCI, ORCAA, Warden AI, Holistic AI, Credo AI, ConductorAI) or academic audit was found on softgarden.com or in the help center (sitemap review plus on-site searches for 'bias' and 'Diskriminierung'), and trust.warden-ai.com/softgarden returns HTTP 404.

    Have a URL that disputes this?

  • Public statementCaptured Sep 25, 2026

    softgarden says its selection AI incorporates 'insights from past successful hiring processes' and therefore 'reduces the risk of discrimination and ensures a fair selection process', but publishes no testing, selection-rate or impact-ratio data to support the claim.

    https://softgarden.com/en/solutions/ai-recruitment/

  • Public statementCaptured Sep 25, 2026

    A March 2024 softgarden blog post says the AI matching also draws on 'wertvolle Erkenntnisse aus früheren erfolgreichen Einstellungen' (insights from earlier successful hires) and that AI 'reduziert Bias', with no audit or evidence offered.

    https://softgarden.com/de/magazin/blogartikel/ki-recruiting-mit-softgarden/

  • DocumentationCaptured Sep 25, 2026

    The Screening Assistant FAQ says it processes only selected CV data and leaves out protected characteristics such as age or gender, an input-exclusion design claim rather than a measured outcome audit.

    https://softgarden.com/en/products/ai-screening-assistant/

FRIA Support

4 items30

Data Governance Disclosure

7 items54

Human Oversight Design

5 items58

Post-Market Monitoring

6 items44

Customer Documentation

3 items58

Editorial notes

Company overview

softgarden e-recruiting GmbH is based in Berlin. It was founded in 2001 in Saarbrücken and has been part of Grupa Pracuj S.A. since 2022. It sells a cloud ATS with multiposting, career sites and onboarding, starting at €199 a month for the Core plan. Its main hiring AI is AI-Assisted Applicant Matching, available to all customers since May 2025 and off by default. A proprietary model compares CVs (parsed by Textkernel) with job requirements and marks applications with a 'particularly high match score' as an 'AI Match', which the German documentation calls top candidates. Recruiters can filter the list to show only matched applicants. A newer LLM-based AI Screening Assistant runs on Microsoft Azure in the EU: it scores every applicant against criteria the AI proposes and recruiters can edit, and highlights supporting CV evidence. Generative features such as job-ad and email drafting run on OpenAI. The security credentials are solid: DEKRA ISO/IEC 27001:2022 certification, German-only hosting, a public DPA and a 2025 penetration-test certificate. The AI-specific documentation is thin.

Regulatory exposure

Recruitment and selection AI is high-risk under EU AI Act Annex III point 4, and the Digital AI Omnibus moved those obligations to 2 December 2027. softgarden says its matching was 'initially categorized as high-risk' but is exempt under Art. 6(3)(a) and (d) because it 'supports decision-making rather than replacing human evaluation'. That claim is weak. Art. 6(3) covers narrow procedural or preparatory tasks that do not materially influence outcomes. This feature scores applicants, flags top candidates and offers a filter that shows only AI-matched applications. softgarden's own materials say it helps recruiters 'filter out' top applicants ('herauszufiltern'), headline it 'Easier Selection with AI Matching', and describe AI used 'when selecting applicants (matching)'. That is the core Annex III 4(a) use case: analysing and filtering applications and evaluating candidates. Art. 6(3)'s final subparagraph also keeps any Annex III system that profiles natural persons high-risk, and automated evaluation of applicants' suitability for a job fits the GDPR Art. 4(4) definition of profiling. Being decision support is not the legal test. softgarden has published no Art. 6(4) assessment. The Screening Assistant, which gives every applicant an overall score, is described only as 'in line with the EU AI Act'. Deployers who rely on the exemption risk under-preparing for their Art. 26 duties (oversight, logging, worker information). US obligations arise only where customers use the features for roles in those jurisdictions: NYC LL 144 audits and notices, Illinois HB 3773 notice and anti-discrimination duties, Colorado SB 26-189 disclosures from 2027, and California's FEHA automated-decision rules. softgarden publishes no US guidance or bias audit.

Path to a higher score

softgarden could start by publishing its Art. 6(4) assessment, including why the profiling carve-back does not apply. The more credible route is to treat Matching and the Screening Assistant as high-risk and publish Annex IV-style documentation: intended use, where the training data comes from (including the past-hire signal and the reuse of customers' anonymised applicant data), accuracy and limitations, and instructions for use, working toward ISO/IEC 42001. It should commission an independent bias audit that publishes selection rates and impact ratios by sex, age and other protected characteristics where lawful, plus an LL 144 audit if US customers use the features. It should also publish Art. 26/27 deployer guidance with a FRIA template, and document how reviewers override AI flags, how AI outputs are logged, and what guardrails apply to bulk actions on screening results. Finally, it should add AI components and model-update notes to its status page and product-update changelog, list the Azure LLM host in the DPA subprocessor annex, and replace the 'Lorem ipsum' placeholder FAQ entries on the Screening Assistant page.

Regulatory frame

What applies to ATS-integrated AI.

Ranking features are inside EU AI Act Annex III §4 (high-risk) and inside NYC Local Law 144 when they affect NYC-based candidates. Customers should treat the AI surface as a distinct deployer-FRIA scope from the rest of the ATS.

Keep reading

All ATS-integrated vendorsHow this score is builtAnnex IV section-by-section: what AI hiring vendors must documentWhat makes a defensible NYC LL 144 bias audit

Compare

softgarden against its nearest-scoring peers.

The closest scores in ATS-integrated AI, side by side.

Build any comparison

Ranking

Others rated in ATS-integrated AI

Ranked 57 of 77 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.

  1. #54Breezy HR43F
  2. #55Paylocity43F
  3. #56Recruitee42F
  4. #57softgardenThis profile42F
  5. #58Zoho Recruit42F
  6. #59Cezanne Recruitment41F
All ATS-integrated vendors

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.