# AI in staffing: use cases and the EU AI Act > AI in staffing and workforce services: which tasks can be automated today, what the EU AI Act means for recruitment, and how we implement it GDPR-compliantly. Canonical: https://www.waterglass.ai/en/ai-in-staffing/ Language: English (Deutsch: https://www.waterglass.ai/de/ki-in-der-personaldienstleistung.md) Company: Waterglass FlexCo, Börseplatz 1/3/6, 1010 Vienna, Austria · hi@waterglass.ai ## AI in staffing Documents, contracts, payroll and matching: we automate the recurring work in staffing and workforce services, inside the EU and with the EU AI Act in view. ## From practice: What AI delivers in staffing today For the German staffing platform [InStaff](https://www.waterglass.ai/en/portfolio/instaff/) we introduced and implemented a GDPR-compliant AI strategy across the operation. More than eight endpoints cover document processing, entirely inside the EU, and a model we trained ourselves on European infrastructure answers in under 0.5 seconds. Where the work used to run almost entirely by hand, 55 to 80 per cent of each process is now handled by AI. The pattern is the same across the sector: every placement brings paperwork with it, from contracts and proofs through to timesheets and invoices. That work rarely needs judgement, but it recurs often enough to scale with the business and tie up people’s time. At the same time, staffing is the area where AI has to look most carefully. Personal data is involved, and selection decisions fall under the EU AI Act. So we draw a firm line: administrative work can be automated, assessing people is a separate matter with separate rules. We work from Vienna across the whole DACH region, on site or remote. [Get in touch](https://www.waterglass.ai/en/contact/) for a free, no-obligation intro call. ## Why Waterglass: What we actually do in staffing ### Documents and proofs Capture contracts, qualification records, timesheets and invoices automatically and write the results back into existing systems. At [InStaff](https://www.waterglass.ai/en/portfolio/instaff/) more than eight endpoints do exactly that. ### Your own model instead of data leaving For [InStaff](https://www.waterglass.ai/en/portfolio/instaff/) we trained a small language model that runs on European infrastructure, answers in under 0.5 seconds and is cost-efficient to operate. ### A clear line around selection decisions We automate the administration around a placement. Where a system filters applications or evaluates people, the EU AI Act applies, and we plan for that from the start. ### Processing in Europe Candidate and employee data do not leave the EU. On request with European-hosted or open models on [infrastructure we run ourselves](https://www.waterglass.ai/en/products/tide/). ## How it runs: Four steps to the first result A sequence that works alongside day-to-day business: few meetings, clear interim results and a decision after every step. ### Intro call In an intro call we establish your starting point and your goals. Then we go through the options for putting AI to work in your company. Free and without obligation. ### Assessment We review processes, systems and data and talk to scheduling, recruiting and payroll. As we go, we record which steps are pure administration and which fall within the scope of the EU AI Act. ### Workshop and prioritisation One workshop with leadership and team puts the use cases in order by value, effort and risk. Afterwards it is clear what to start with and who owns it internally. ### Pilot and operations The first use case goes into production, with real data and a measurable result. On request we then take on running it, so the solution does not fail for lack of staff. ## Selected references: Successful client projects > "AI agents are a transformative technology that unlocks efficiencies across a wide range of businesses." > > — Silia Traub, Director AI & Strategy, willhaben > "We are using AI agents to radically improve our business efficiency and quality of service for our customers." > > — Oliver Lukesch, CTO, StartMatch ## We integrate seamlessly into your infrastructure Slack, Microsoft Teams, Outlook, Gmail, Notion, Google Drive, Salesforce, HubSpot, OpenAI, Anthropic, Datadog, Jira, Zapier, Stripe, GitHub, Linear, QuickBooks, Zendesk, Asana, Intercom, Google Calendar ## Questions & answers: Common questions about AI in staffing **Which tasks can be automated first?** The administrative work around a placement: capturing contracts and proofs, checking timesheets, preparing invoices, answering standard enquiries, moving data between systems. At [InStaff](https://www.waterglass.ai/en/portfolio/instaff/), 55 to 80 per cent of each process is handled by AI today. **Does the EU AI Act allow AI in recruitment?** Yes, but with obligations attached. Annex III of the EU AI Act classifies as high-risk those AI systems used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, to analyse and filter job applications, and to evaluate candidates. High-risk does not mean banned, it means bound to duties such as risk management, documentation, logging and human oversight. **When do those obligations start?** For Annex III high-risk systems, as things currently stand, from 2 December 2027. That date was moved relative to the original text of the Regulation, which is why many articles online still quote an earlier one. Other parts already apply: since 2 February 2025 the Article 5 prohibitions, including emotion recognition in the workplace, and the Article 4 AI literacy duty, which binds providers and deployers alike. **Is this legal advice?** No. We build the systems and the evidence trail that goes with them: documentation, logging, and clear boundaries where a human decides. The legal assessment of your specific use belongs with your legal counsel, and we are glad to work alongside them. **How do you handle candidate data?** It stays in the EU. We design data protection in from the start: processing inside the EU, data minimisation and clear rules about which fields a model gets to see at all. For [InStaff](https://www.waterglass.ai/en/portfolio/instaff/) the core runs on a custom model on [European infrastructure](https://www.waterglass.ai/en/products/tide/). **Does this work with our applicant tracking system or ERP?** Usually yes. What matters is not the vendor but whether data can be pulled out of the system and results written back, through exports or a database connection if need be. Where an interface is missing, we say so in the assessment rather than adding it later as extra scope. **What does an AI project in staffing cost?** Scope depends on your starting point and the question, from a compact assessment through to full implementation with operations. After a short intro call you get a fixed, no-obligation quote, scope, sequence and cost included. **Who runs the solution afterwards?** We do, if you want. Running it is a separate, clearly priced part of the offer: monitoring, updates and adjustments when your workflows change. If you want to run it in-house, we document accordingly and hand over. ## Let us talk about AI in your staffing business Tell us where your company stands. We will show you where AI has the greatest leverage and put together a free, no-obligation quote. ### How can we help? Book an intro call and we will look together at which workflows are worth it and which are not.