Note: The "AI Screening Assistant" feature is currently in a beta phase and is therefore not yet available to all customers. Once the feature has been fully rolled out, we will provide an update in the "News" section, accessible via the chat bubble in the bottom-right corner of the system.
What the AI Screening Assistant can do:
Automatically derives screening criteria from job descriptions and screening questions.
Automatically evaluates every application upon receipt—no manual trigger required.
Provides transparent evidence: Each evaluation highlights the exact passage from the CV.
Allows for manual review and overrides: Any AI-generated evaluation can be corrected.
Includes a bias check for all manually created criteria.
Facilitates high-volume triage: Sort, filter, and perform bulk actions.
Ensures team-wide visibility of any corrections.
GDPR-compliant; does not use protected characteristics.
Hosted in the EU; data is not used for AI training.
Not a "black box"—every evaluation is fully traceable.
See at a glance exactly why a candidate is a match!
How the Assistant helps you
The Assistant automatically suggests relevant screening criteria, which you can freely adjust or supplement. It then compares each application against these criteria and shows you:
how many criteria have been met (e.g., 4 out of 5)
for each criterion:
full match
partial match
no match
Insufficient data
In the detailed view, the relevant sections of the resume are highlighted—allowing you to see at a glance why a match exists.
You always retain full control: you can review and correct every assessment and decide how to proceed with each application.
Security, Transparency, and Compliance
Processing of only relevant data from the resume
Protected characteristics (age, gender, nationality, etc.) are not taken into account
Language model runs on Microsoft Azure servers within the EU
Data is neither stored nor used for AI training
Usage in compliance with the GDPR and the EU AI Act
AISA does not make any decisions – every action is initiated by you
Setting up the AI Screening Assistant
Path: Settings → AI Features
When the AI Screening Assistant is activated, this setting is automatically inherited by all configured sub-tenants. It can be disabled for individual sub-tenants at any time.
On the same settings page, you can grant or revoke user access at any time, allowing users to edit evaluations and view only those criteria that have been fully met.
Alternatively, you can also activate the feature directly within an existing job under the "Applications" tab.
To do so, click the "AI Screening Assistant" button and then toggle the "Activate" switch.
Once activated, all candidates applying to your company will be shown the updated privacy policy, which permits the processing of their data by the AI.
How the AI Screening Assistant works
The AI Screening Assistant supports you through five clear steps—from defining criteria to making the final decision.
1. Defining screening criteria
The Assistant automatically generates relevant criteria based on the job description. You can:
edit or delete criteria
add up to 10 custom criteria
formulate criteria using natural language
include detailed prompt instructions
add requirements that are not explicitly stated in the job posting
have your criteria checked for clarity, bias, and compliance with anti-discrimination laws (such as the German General Act on Equal Treatment, AGG)
receive warnings and alternative suggestions for potentially problematic phrasing (e.g., "native speaker").
2. Automatically evaluate applications
Once the criteria have been established, the Assistant automatically evaluates every incoming application.
You can see at a glance:
how many criteria have been met
for each criterion:
fully met
partially met
not met
insufficient data
The AI Screening Assistant uses the following for this process:
Parsed CVs
Screening questions
The job description
3. Review candidates in detail
In the detailed view, you can see:
Criteria and the resume side by side
Highlighted text segments as evidence
Indicators for missing information
You can:
manually review each evaluation
override any evaluation
optionally add a justification
make all decisions yourself
Corrections are visible to the entire team.
4. Sort, filter, iterate
You can filter applications based on:
must-have criteria
complete, partial, or unknown matches
top matches (≥ 80% met)
weak matches (≤ x% met)
At any time, you can:
adjust criteria
re-run the assistant
update the results
5. Full integration into the softgarden ATS
You can:
Combine AI filters with existing ATS filters
Select multiple applications at once
Perform bulk actions (forward, move, send messages)
Everything takes place within the familiar softgarden system—no third-party tools required.
Important Information Regarding the AI Screening Assistant (FAQ)
The AI Screening Assistant offers numerous features designed to help recruiters evaluate job applications efficiently and in compliance with legal requirements. The following points address key questions concerning its use, the legal framework, and technical details.
Use for Applications Already Received
From a technical standpoint, the AI Screening Assistant can be applied to applications that have already been submitted. However, this is not recommended from a legal perspective, as candidates must be informed before their data is processed by AI. When the AI Screening Assistant is activated, a corresponding notice is automatically added to the company’s privacy policy.
Recommendation: Use the AI Screening Assistant only for job openings created after the feature has been activated.
Documentation of Changes
AI-judgment overrides: These are fully logged and displayed transparently.
Manually added screening criteria: These are clearly labeled as "Manually added."
Compliance with the General Act on Equal Treatment (AGG)
The AI Screening Assistant blocks screening criteria that violate Section 1 of the General Act on Equal Treatment (AGG)—for example, criteria related to nationality, ethnicity, religion, gender, age, or sexual identity. In such cases, the AI Screening Assistant prompts the user to provide further clarification. The user can then decide whether to adjust the criterion or keep it as is.
Minimum Requirements for Generating Screening Criteria
The AI Screening Assistant requires at least the job title to generate screening criteria. A job description is helpful but not strictly mandatory.
Usage Without CV Parsing (Textkernel)
Usage without CV parsing is not recommended, as essential data for the screening process would be missing.
Filtering Logic
Users can filter and sort flexibly, based on criteria such as:
Must-have criteria
Full, partial, or unknown matches
Top matches (≥ 80% of criteria met)
Weak matches (≤ x% of criteria met)
Data Privacy Policy
When the AI Screening Assistant is activated, a notice is automatically added to inform candidates that their data is being processed by AI.
Hosting & Data Processing
The AI Screening Assistant uses a Large Language Model (LLM) from OpenAI, hosted on Microsoft Azure within the EU. Data is neither stored nor used for training purposes.
Language
The AI Screening Assistant operates in the language configured for the specific job.
Creating Screening Criteria Later
Screening criteria can be created or edited at any time: Job → Qualifications → Candidate Criteria
Roles & Permissions
Creating and editing screening criteria is restricted to the Recruiter role.
Calculating the number of matches
The AI Screening Assistant counts only complete matches (e.g., 3/5). Partial matches are not included in the count.
Disabling the override function for evaluations
Administrators can disable the option to override AI-generated assessments.
Number of applications displayed
Up to 25 applications are displayed per page.
Definition of “Top matches” / “Weak matches”
Top matches: ≥ 80% of criteria met
Weak matches: ≤ x% of criteria met