Frequent
The task happens often enough that a better method will matter.
Practical guide · AI in HR
How should HR teams use AI responsibly?
Start with preparation and communication work, use approved sources and tools, define a human reviewer, and never delegate employment decisions to a model.
- George Raymond-Alshoufi
- Updated 16.08.2026
- 8 min read
The short answer
The best first AI use cases in HR are frequent, language-heavy and easy for a person to verify. A good workflow has an approved source, a clear output, a named reviewer and no automated decision about an individual.
What does AI in HR mean?
AI in HR means using systems such as Microsoft 365 Copilot, ChatGPT Enterprise or approved internal assistants to support People work.
Useful support includes structuring information, preparing drafts, summarising approved material, generating alternatives and turning repeated tasks into reviewable workflows.
It should not mean handing hiring, promotion, performance, compensation or termination decisions to a general-purpose AI tool.
Choose the right starting point
A four-part test for an HR use case.
If one criterion fails, reduce the scope or choose a safer task.
Grounded
The work can be based on approved policies, templates or source material.
Reviewable
A responsible person can check accuracy, omissions, tone and impact.
Non-decisional
The output supports preparation; it does not determine an outcome for a person.
Good first AI use cases for HR
Policy Q&A preparation
Useful patternDraft an answer from approved policy material and show the source passages used.
Control to addHR reviews exceptions, local context and the final message.
Role brief and job draft
Useful patternTurn a structured role intake into a clearer first draft and challenge vague requirements.
Control to addReview inclusion, necessity of criteria and employment-market context.
Interview preparation
Useful patternCreate role-related questions and a consistent note-taking structure.
Control to addDo not score candidates with a general model; people assess evidence.
Onboarding plan
Useful patternSequence approved content, stakeholders and first-month learning goals for a role.
Control to addThe manager adapts the plan to the person and actual team needs.
Meeting follow-up
Useful patternTurn notes into decisions, questions, actions and owners with a source trail.
Control to addThe meeting owner confirms what was agreed before circulation.
Learning support
Useful patternCreate practice exercises and reflection questions from approved capability goals.
Control to addValidate accessibility, relevance and whether outcomes can be observed.
Human accountability
Decisions that should remain human
AI can help organise evidence or prepare questions, but accountability cannot be transferred to a model. High-impact decisions require context, due process, explanation and a responsible human decision-maker.
- Hiring and rejection
- Promotion and succession
- Performance ratings
- Compensation
- Disciplinary action
- Termination
A responsible 30-day pilot
Define
Select one workflow, source set, owner, reviewer and data boundary.
Test
Run realistic examples, document failure modes and refine the instructions.
Compare
Compare the new method with the existing process using agreed quality signals.
Decide
Keep, change or stop the pilot. Record the approved workflow and ownership.
Measure more than time saved
- Factual accuracy
- Completeness
- Consistency
- Review effort
- Sensitive-data handling
- User confidence
- Escalations and exceptions
AI for HR · Switzerland
Turn the guide into a working team method.
The AI for HR programme applies these principles to your tools, policies and recurring workflows.