Recruitment
AI in recruitment: which ATS features really help with hiring?
Which AI features in ATS software really support recruitment? See applications for job ad texts, summaries, scheduling and selection criteria.
AI in recruitment can help with job adverts and processing information. The quality of a proposal and the checking of selection still need attention. See which ATS features you can investigate and which prompts suit a first demonstration.
The role of an ATS and the arrival of AI assistance
An Applicant Tracking System (ATS) acts as the central platform for structuring recruitment campaigns, managing vacancies and following up candidates. As the supplier Personio explains in its overview of ATS systems and selection criteria, the main strength of an ATS lies in its overview of the recruitment pipeline and a structured flow of communication with applicants.
The introduction of AI in ATS software focuses mainly on language processing and pattern recognition. Suppliers build in digital assistants that speed up routine text-based tasks. A documented example is the Personio Assistant, with which HR staff can retrieve reporting data through targeted questions, where the supplier describes that access rights are taken into account. This example concerns HR reporting and does not in itself prove AI selection of applicants. It is essential here to make a sharp distinction between text support on the one hand and substantive selection decisions on the other.
Four practical uses of AI in recruitment
In modern recruitment processes, AI can support four time-consuming tasks:
- Writing drafts of job adverts: Based on keywords about the role, the experience required and the education level, the software generates a first draft. This saves time when starting a vacancy, although manual adjustment to your own company culture and terms of employment remains necessary.
- Structuring and summarising CVs: The software extracts work experience, language skills and education from unstructured documents such as PDF files and places them in fixed fields within the candidate profile in a structured way.
- Draft messages for candidate communication: The system drafts proposals for acknowledgements of receipt, invitations to interviews or courteous rejections. The recruiter adjusts the text before it is actually sent.
- Support with interview scheduling (not AI by default): Scheduling modules compare interviewers' calendar availability and propose suitable time slots to candidates.
Why automated rejection is undesirable
Some discussions about recruitment technology suggest that AI can assess or reject candidates independently. In a responsible recruitment practice, however, automatic exclusion is strongly discouraged:
- Risk of algorithmic bias: Language models rely on historical patterns. If earlier personnel data over-represented certain characteristics, an algorithm can unwittingly adopt these preferences and disadvantage qualified candidates.
- Errors in automatic text analysis: Unusual layouts, creative job titles or missing keywords can cause the software to overlook relevant skills. A candidate who is an excellent match in substance but happens to choose different wording could wrongly be passed over.
- Privacy and data protection: Application data contains sensitive personal information. For a first trial, use fictitious CVs and interview notes. Have real use assessed in advance within your organisation's privacy policy. For the rules that apply to privacy and automated decision-making, always consult the official guidelines of supervisory authorities such as the Dutch Data Protection Authority (Autoriteit Persoonsgegevens).
The principle of human oversight therefore remains leading: AI can make administrative proposals and organise data, but a recruiter or hiring manager always personally assesses a candidate's suitability.
Fictitious practical example: prompt for a job advert
To show how you can use AI in practice for recruitment texts without sharing personal data, the box below shows a structured example prompt:
Fictitious example prompt for drafting a job advert:
"Write a professional and inviting job advert for a Junior Supply Chain Planner (36 to 40 hours, based in Utrecht). Education level: university of applied sciences (hbo) in Logistics, Business Administration or similar. Knowledge of inventory management and experience with ERP software are a plus. Tone: clear, direct and enthusiastic. Avoid clichés such as 'jack of all trades' and do not invent terms of employment. Mark missing information about salary, working hours and training budget as still to be determined."
After generating the text, the recruiter checks the content: do the primary and secondary terms of employment match organisational policy, and is the text inclusive in its wording? Only after this human editing step is the vacancy published.
Comparing ATS software and suppliers
In the recruitment software category, organisations choose from various platforms:
- Personio: Offers a broad platform in which vacancy management, candidate communication and selection routes are directly connected to the central HR administration and onboarding.
- Cegid HR: Offers talent management and recruitment solutions for organisations with complex selection procedures, multiple assessment teams and specific job groups.
When exploring software, it is advisable to ask suppliers which AI features are actually available within the chosen licence type, which underlying language models are used and how data security is arranged contractually.
Possible AI features and selection criteria in a demonstration
The table below shows features you can investigate during a demonstration of an ATS:
| Recruitment activity | Possible AI support (to be verified) | Practical point of attention in a demo | Human oversight required |
|---|---|---|---|
| Job adverts | Generating a draft based on keywords. | Does the software check for inclusive language and the desired tone of voice? | Editorial review of terms of employment and culture. |
| CV parsing | Reading education and work experience into data fields. | How reliably does the parser recognise tables and different file formats? | Checking the original document if there is doubt about experience. |
| Candidate communication | Drafting messages for status updates and rejections. | Can templates easily be personalised per candidate? | Personal review and approval before sending. |
| Interview scheduling | Proposing appointment times based on calendars. | Does the calendar integration work with your existing email environment? | Checking that the right interviewers have been selected. |
| Candidate selection | Do not apply automatic selection or rejection. | Make sure no automatic exclusion filters are active. | Fully human assessment of suitability. |
Selection advice for recruitment teams
When choosing an ATS, do not be guided solely by promising marketing terms around AI. First examine the basic quality of the system: a clear candidate pipeline, user-friendly communication tools and reliable reports form the foundation of successful recruitment. Regard AI features as handy aids for draft texts and data entry, while responsibility for assessing candidates always remains with your recruitment team.
Check whether a summary leaves out important information
Make two fictitious CVs with the same relevant experience but a different layout. Put the education at the top in one and at the bottom in the other. Ask the assistant only to summarise the experience and education mentioned, with references to the original passages. Leave missing data as unknown.
Compare the results: is the same experience recognised, do gaps in the information become visible and do no invented qualifications appear? Then have a recruiter assess the original documents. A summary can reduce searching; the quality of that summary is not a ranking of candidates.
