HR's AI training has to be different, because HR's decisions affect people's jobs. Here's what to teach, where the legal lines sit, and how to structure it.
TL;DR
HR teams are often handed two AI jobs at once. They're expected to use AI in their own work, and they're often asked to lead AI training for everyone else. Both matter, but the first comes first. It's hard to lead a programme you haven't been through yourself.
This piece covers the first job: what HR teams need to learn about AI, where the risks are higher than in other functions, and how to structure the training. For the organisation-wide side, our guide to AI strategy for HR leaders is the better starting point.
A marketing team that gets AI slightly wrong publishes a weak blog post. An HR team that gets it wrong can screen out qualified candidates, make a decision about someone's job based on flawed output, or mishandle personal information. The stakes sit closer to people's lives.
Regulation reflects that. The EU AI Act's Annex III lists employment and workers' management among its high-risk areas, including AI used to place targeted job adverts, filter applications, and evaluate candidates, as well as AI used in decisions on promotion, termination, task allocation, and monitoring performance. The European Commission's AI literacy guidance notes that organisations deploying high-risk systems must make sure the staff using them are trained to provide human oversight.
In South Africa, section 71 of POPIA says a person may not be subject to a decision with legal consequences, or that affects them substantially, based solely on automated processing intended to profile them, including their performance at work. There are exceptions, so get legal advice on your specific use, but the direction is clear: a person needs to stay meaningfully involved.
The training should be built around HR's actual work. Here's how I'd map it.
| HR activity | Where AI helps | Where a person must decide |
|---|---|---|
| Recruitment | Drafting job adverts, interview question banks, candidate communications | Shortlisting, assessment, and every hiring decision |
| Onboarding | Welcome packs, first-week plans, FAQ drafts | What's promised to the new hire |
| Policy and communication | First drafts of policies, plain-language rewrites, summaries of legislation | Final wording and legal sign-off |
| Employee queries | Draft answers from approved policy documents | Sensitive, personal, or disciplinary matters |
| Learning design | Course outlines, scenarios, quizzes | What capability the business needs |
| People analytics | Summarising survey comments, spotting themes | Conclusions about individuals |
The pattern is consistent. AI handles drafting, summarising, and first-pass analysis. People make the decisions that affect someone's job. Training should practise both halves: using AI well on the first, and knowing exactly where the line sits on the second.
I'd include at least one exercise where the group compares an AI-drafted job advert against the original and edits it for language that might put off certain applicants. It teaches drafting, judgement, and bias awareness in one go.
Bias. Models reflect patterns in the data they learned from, and those patterns can disadvantage groups of people. HR teams should practise spotting biased output in adverts, summaries, and assessments, and know that a tool's confidence says nothing about its fairness.
Personal information. Which tools are approved for employee and candidate data, and which aren't. Pasting a CV into a consumer AI tool may put you in breach of your data protection obligations. Make the rule simple and specific.
Automated decisions. Where the law draws lines, as in POPIA section 71 and the EU AI Act's high-risk categories, and what "a person in the loop" means in practice. In my view, glancing at an AI shortlist before approving it falls well short of meaningful involvement.
Transparency. What candidates and employees are told about AI use in processes that affect them. Being upfront builds trust, and depending on where you operate, it may be required.
Vendor questions. Many HR systems now include AI features. HR teams should know what to ask: what the feature does, what data it uses, how it was tested for bias, and what the vendor's obligations are. Our piece on AI in recruitment covers this for hiring tools specifically.
A baseline for the whole HR team. How the tools work, where they fail, the organisation's rules, and hands-on practice on drafting and summarising tasks. Half a day covers it.
A deeper module for higher-risk roles. Recruiters, HR business partners handling performance and disciplinary matters, and anyone working with people analytics need the full risk module, practised on realistic scenarios.
Practice on anonymised material. Use real HR documents with personal information removed. A real policy or a real job advert teaches more than a generic example, and anonymising it is a useful lesson in itself.
Written guidance to take away. A one-page summary of approved tools, what data can go where, and which decisions always need a person. It's what people will check when they're unsure.
For individuals who want more structure, there are HR-specific options. SHRM offers an AI+HI specialty credential aimed at HR professionals. Coursera hosts a Generative AI for HR specialisation covering recruitment, onboarding, training, and performance management.
These work well for personal development. They won't cover your organisation's tools, policies, or legal context, so pair them with an in-house session on how your team does things.
Once the HR team is confident, it's in a far better position to design and run organisation-wide training. It understands the tools, the risks, and what it feels like to change your own way of working.
That experience helps with the hardest part of AI adoption, which is behaviour change rather than knowledge. Our guide to AI upskilling covers how to run that wider programme.
For individuals, HR-specific options include SHRM's AI+HI specialty credential and Coursera's Generative AI for HR specialisation. For a team, an in-house session using your own policies, tools, and legal context will change more.
For drafting job adverts and policies, creating onboarding and learning content, summarising survey comments, and preparing answers to employee queries from approved documents. Decisions about individuals should stay with people.
Yes. Annex III lists AI used for recruitment and selection, including filtering applications and evaluating candidates, as high-risk, along with AI used in promotion, termination, and performance decisions.
POPIA section 71 restricts decisions based solely on automated processing that significantly affect a person, with some exceptions. Keep a person meaningfully involved in decisions and get legal advice on your specific use.
Recruiters, HR business partners handling performance or disciplinary matters, and anyone working with people analytics, because their AI use touches decisions about individuals.
If you want to see how ready your people function is before planning training, our free AI Readiness Diagnostic gives you a scored view across strategy, data, people, process, and tooling.
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic