AI governance in HR and the transformation roadmap
Step 1 / 4·Governance: five questions you must answer
0/4 steps done
Learning objectives — after this module you will:
- Build an AI governance frame for the HR function
- Apply data principles and human-in-the-loop rules
- Build a defensible 12-month transformation roadmap
Governance: five questions you must answer
Lay the governance foundation before scaling.
AI governance is often seen as the thing that slows everything down. The opposite is true: missing governance is why AI programmes get halted abruptly at the first incident. A lightweight governance frame lets you scale FASTER, because you don't have to re-seek permission for every use case.
The five questions and their minimum answers
| Question | What to specify |
|---|---|
| Which data may go into AI? | A clear list of what must NOT: identified pay data, health records, disciplinary information, personal identifiers |
| Who may use AI for what? | Use cases grouped by risk level; high-risk ones need approval |
| When is human review mandatory? | Any decision affecting opportunity or livelihood: selection, appraisal, discipline, termination |
| Do employees get to know? | Transparency about where AI is used in processes that concern them |
| Who is accountable when it goes wrong? | The authorised decision-maker remains accountable — 'the AI suggested it' is not a valid explanation |
The non-negotiable principle
AI does NOT make decisions about people. It can order, summarise, suggest and draft — but who gets hired, how someone is rated, and how discipline is handled must be decided by an authorised human who carries the accountability. That's both an ethical principle and, increasingly, a legal requirement.
Key takeaway: Five governance questions; the immovable principle is that AI never decides about people.