AI Training for HR Teams 2026: Program, Boundaries and Price (from €1,200 net per day)
HR is the role where AI is used most and regulated hardest. Good AI training for HR teaches one thing above prompts: where assistive use ends (drafts, summaries, candidate communication) and high-risk territory begins (candidate ranking and selection) under Annex III of the AI Act. At Syntalith we run it from €1,200 net per day, for a group of up to ~15 people.
AI training for HR teaches one thing more important than prompts: where assistive use ends and high-risk territory begins. A model can draft job ads, summarize surveys and prepare materials, but ranking and selecting candidates are high-risk systems under Annex III of the AI Act. We run closed AI (Claude) training for HR from €1,200 net per day, for a group of up to ~15 people.
Quick answer
HR is a special role to train on AI because it combines two things at once: this is where AI is used most (ads, communication, documents) and where the rules bite hardest (recruitment and employee evaluation). So good training is not a day of prompt tricks, it is learning to work within boundaries. At Syntalith we price it plainly, net:
- free process scan (€0): a 30-minute engineer call plus a written takeaway in two business days, to establish what the HR team actually needs before you buy anything,
- AI (Claude) training for HR (from €1,200 net per day): a closed workshop on your real tasks, one or two days, online or on-site, for one workshop group (up to ~15 people),
- at the end: a written way of working with AI, the rules for candidate data, plus materials and an attendance list for AI Act article 4.
The daily rate usually sits in the €1,200–1,500 net band, and you get the final per-scope quote after one call. The full price list is on the pricing page, and the team offer is described on the AI training for teams page. This piece is about the HR team's competence; if you are looking for a ready system to organize CVs, we describe it separately in the piece on an AI agent for screening 200 resumes.
How to structure AI training for HR so it changes work
Changed work does not come from one day in a room, it comes from three steps around it, the same as in AI training for teams. The difference in HR is in the examples: we work on the tasks the department really does, and on their boundaries from the start.
1. Audit what HR really does. Before anyone stands in front of a group, you need to know where the department spends its time: how many ads it writes a month, how many candidate messages go out by hand, what onboarding looks like, where the surveys and exit interviews sit that need summarizing. Without this the program teaches generic things. For us this step lives in the call about the team and the goal, from which the scope and examples are drawn.
2. A closed workshop on real HR tasks. The day goes on the department's work moved onto the model, with the output verified on the spot: a specific job ad, a real candidate message, an onboarding document. Across two days, the first goes on hands-on work with the tool, and the second on what matters most in HR: the boundaries of what may be handed to AI, the rules for personal data, and the decision trail.
3. Follow-up and a written policy. What remains after the workshop is a document: what may be pasted into the model, what may not, where the process must stop, how to leave a trail, and who approves. It is that artifact, not the day itself, that decides whether HR works more safely on Monday. Without it, training is theater.
What AI in HR may do on its own, and what it may not
This is the heart of the whole training and the main reason an HR team should learn AI differently from the rest of the company. The line runs between assistive use (the model prepares, a human approves) and decision-shaping use (the model influences a decision about a person). The latter, in recruitment and worker management, is classified as high-risk under Annex III of the AI Act, and obligations for such systems phase in with transition periods running to December 2027. We state this generally and without panic: it is not legal advice, only a reason for the team to be able to tell which side of the line it stands on.
The four most common HR tasks and their boundary:
| HR task | Assistive use (allowed, with review) | Decision-shaping use (high-risk territory) |
|---|---|---|
| Job ads and candidate communication | ad drafts, message variants, clarifying requirements, confirmations and invitations after a human decision | automatic rejection of candidates, sending refusals without review, hiding applications from the recruiter |
| Onboarding materials and internal documents | drafts of materials, simplifying procedures, draft policies and checklists | binding HR decisions derived from a document without a human |
| Summaries (exit interviews, surveys) | summarizing and organizing responses with anonymization, surfacing topics to discuss | scoring or ranking people from summaries, personnel conclusions without verification |
| Preparing training and policies | a training plan, a draft policy, educational materials for the team | treating a generated policy as final and binding without legal review |
The left column is work the workshop genuinely speeds up, provided a human stays in it to approve the output. The right column is not a one-day training task. Ranking, assessing and selecting candidates and evaluating employees is a high-risk system deployment with the full regime: human oversight, data quality, logs, risk management and audit. It is usually not something an HR team should build on its own after a day of workshop. Good training teaches you to see that line, not to cross it. How to design the system itself on the right side of the line, we break down in the piece on an AI agent for CV screening.
Candidate and employee data: the hard rule
We repeat this one rule on the workshop until it becomes a habit: candidate and employee data (CVs, appraisals, medical data) does not go into unapproved AI tools. GDPR applies no matter how convenient a public chat is, and some of that data falls into special categories that need extra caution and a separate legal basis. Pasting a CV into a random tool to "quickly summarize" it is not a shortcut, it is moving personal data outside the company's control.
Training turns that rule into specifics: which tool is approved at your company and on what terms, what may be processed in it, how to anonymize survey and exit-interview responses before they reach the model, and where the process must stop when someone is unsure. If the company does not yet have written rules, it is worth setting them in parallel; how to do that we describe in the guide to a company AI policy and employee rules.
What AI training for HR costs (price per day)
A closed AI (Claude) training for an HR team starts from €1,200 net per day. The price is per day and per workshop group (up to ~15 people), not per person. For an HR department of a dozen-odd people a workshop day works out cheaper per head than sending each person to a separate open course, and it teaches on your ads, candidates and documents rather than on room examples.
Three things move the price: the number of days (a one-day introduction or a two-day workshop with the second day on boundaries, data and the standard), the number of groups (a larger team is split into groups of up to ~15 and priced in one offer), and the format (online or on-site). The daily rate usually sits in the €1,200–1,500 net band; the final per-scope figure comes after one call. At the end you get materials, an attendance list and a named certificate that slot into your article 4 competence documentation. What article 4 actually requires and what it does not, we unpack in the piece on AI Act article 4 literacy training.
When NOT to book AI training for HR
Honestly: there are situations where training is a bad buy or a bad first step, however fashionable the topic.
- The goal is "AI should sift CVs for us." That is not a workshop topic. It is a conversation about deploying a high-risk system under Annex III of the AI Act, with oversight, audit and the full regime, and usually not something an HR team should build on its own. Training will not turn manual screening into automatic selection without that conversation.
- HR has no process for candidate data today. If no one knows where the CVs sit, who has access and when they are deleted, set that process first and train on tools afterward. Otherwise the workshop gives the team practice in sending personal data where it should not go.
- A free webinar or the docs are enough. If the goal is a one-off with no real task behind it, start with the free materials from the tool vendors. Paid training makes sense when the department has a process on the table.
It is also worth seeing the wider picture. Statistics Poland (GUS), in data compiled by the Polish Economic Institute (December 2025), reports that 8.7% of Polish companies used AI in 2025, and EY Poland, in an April 2026 report covering 497 medium and large firms, reports that about half cite disappointment or an incomplete return on AI. In HR that common denominator rarely sits in the model. It is usually the lack of competence and rules among the people meant to use AI, and that is exactly the gap good training closes, provided it is structured as a process rather than an event.
How to start
The cheapest sensible first step is to establish what the HR team actually needs, not to buy a training day straight away.
- Book a free process scan and tell us who in HR uses AI or is meant to start.
- Prepare: which tasks the department does each week (ads, communication, onboarding, surveys), which tools it already has, and whether rules for candidate data exist.
- After the call you get a recommendation: a closed workshop on your tasks, help ordering the data rules, or an honest "this is a system deployment topic, not a training one."
Book a free process scan | AI training for teams | See pricing
FAQ
How much does AI training for HR cost?
A closed AI (Claude) training for an HR team at Syntalith starts from €1,200 net per day, typically €1,200–1,500, priced per scope after one call. The price is per day and per workshop group (up to ~15 people), not per person. The program sits on your real HR tasks: job ads, candidate communication, onboarding and documents, and what remains is a written way of working plus documentation for AI Act article 4.
What may AI be used for in HR, and what not?
Assistive use is fine under human review: drafts of job ads and candidate messages, onboarding materials, summaries of surveys and exit interviews with anonymization, preparing training and policies. Decision-shaping use is high-risk territory under Annex III of the AI Act: ranking, assessing and selecting candidates and evaluating employees. That is not a workshop task but a high-risk system deployment with oversight and audit, and usually not something an HR team should build on its own.
Is recruitment AI a high-risk system?
AI used for recruitment and worker management, including CV screening and candidate ranking, is classified as high-risk under Annex III of the AI Act. Obligations for these systems phase in with transition periods running to December 2027. Training does not remove that: it teaches the team to recognize when a tool crosses the line out of assistive use. This is not legal advice, and we state the boundaries generally.
What about candidate and employee data?
The rule is hard: candidate and employee data (CVs, appraisals, medical data) does not go into unapproved AI tools. GDPR applies, and some of that data falls into special categories that need extra caution. The training teaches what may be pasted into a model, what may not, and where the process must stop before someone accidentally sends personal data into a public chat.
When is AI training for HR a bad buy?
When the goal is "AI should sift CVs for us": that is a conversation about deploying a high-risk system with audit and oversight, not a team workshop. And when HR has no process for candidate data today: then set that process first and train on tools afterward. Paid training makes sense when the team has real tasks and wants to run them more safely.