Juliette's Learning Tribe
AI training for HR teams: what to change first, task by task
Start with job descriptions and policies, where AI helps most and risks least. Keep personal data out, and leave decisions about people till last.
Start with the job description. It is the HR document AI helps with most and risks least, because nobody’s personal data is in it. Then policies, then onboarding. Leave anything that decides something about a person until last.
That is the order I would train an HR team in. One task at a time, on your own work, with a person checking every line.
HR is usually the team asked to roll AI out to everybody else. Often before it has had a single afternoon to learn it for itself. So start with your own work.

Job descriptions
Give the tool your rough notes. The purpose of the role, what the person will actually do in a week, who they work with. Ask for a first draft in plain English.
It will be fast and it will be fluent. Then check it.
- Does it describe the job you have, or the job every other company advertises?
- Is there a requirement nobody needs? A degree for a role that never needed one, or ten years for a job somebody could learn in one.
- Is every duty real? AI adds plausible ones.
If you want a starting point that is not a blank page, our catalogue holds UK job descriptions. You can browse the HR roles and see what each one involves.
Policies
Policies are where AI earns its keep. Paste in a policy nobody reads, with names taken out, and ask it to rewrite it so a new starter could follow it. Ask what a manager would still not understand.
Then the person who owns the policy checks every rule. AI will invent a rule that sounds right. It will also write as if you are a big company with a big HR team, when you might be three people.
If your first policy is about AI itself, I wrote about what goes into an AI use policy for employees.
Onboarding
Ask for a first-week plan for a role. Then ask for the questions a new starter is too nervous to ask on day one. That second list is where it gets interesting.
A person still checks the names, the dates, the systems and the building. AI will cheerfully book somebody into an induction that does not exist.
Employee questions
This is where the care starts.
Use AI to draft answers to the questions you get every week. Holiday, payslips, how to book training. Write them once, check them, and keep the good ones.
What you never do is paste in the actual person. Their name, their sickness record, their grievance. The UK’s National Cyber Security Centre says not to include sensitive information in queries to public AI tools. Health information is special category data. The ICO says it needs a high level of security, and should not normally go beyond the people who need it for their jobs. A public chatbot is not one of those people.
Use only the tools your organisation has approved. If none is approved yet, that is the first job, and it is yours.
Analysis
Survey comments are a good test. Give it a few hundred anonymous comments and ask for the themes. It will find them in seconds, and that is still a small marvel to me.
Then read a sample of the raw comments yourself. Check every theme has comments behind it. Watch small groups too. In a team of four, a summary can point at the person who wrote it.
Decisions about people
Recruitment screening, performance ratings, who gets the promotion. Leave these until your team has done everything above for months.
The ICO looked at automated decisions in hiring this year. It is not banning the tools. But it says employers should test for bias regularly and tell candidates how to challenge a decision and ask for a human to review it. If a tool decides and nobody can explain why, you are not ready.
This is what it means in practice, not legal advice. Talk to your data protection lead before any tool makes a decision about a person.
The weak spot
AI writes the average.
Ask it for a values statement and you get everybody’s values statement. Ask for a leadership programme and you get the one built for the middle of the room. I have built plenty of those myself, and some of it was a rubbish bit of work.
So the HR skill is not getting AI to write. It is knowing what your organisation actually needs, and spotting when the draft has gone generic.
What is the best AI course for HR?
The one that uses your own work. A course about the tool shows you where the buttons are. Then you go back to your desk, open a policy, and nothing has changed.
What changes things is doing one real task with AI, checking it, and doing it again next week. I wrote more about that in how to learn AI skills for work.
Where we come in
We built Future Fluency for this. It covers three things: working with AI in your role, managing AI in your role, and managing the change in your role. The lessons use your role and your sector, so an HR adviser learns on HR work.
Your first module is free, with a free account and no card.
If you want to try it with your team first, see how Zavmo works for organisations. There is a 14-day pilot for up to five people, no card needed. Pick one task from this list and bring it.