The People function should be irreplaceable, not indispensable.

AI should own the repeatable. HR should own the irreducible.

Any HR task that produces the same outcome regardless of who is involved — onboarding workflows, offboarding checklists, policy documentation, compliance tracking — should be AI-native. The goal isn't efficiency for its own sake. It's to free up HR capacity for the work that genuinely requires human judgment: conflict, ambiguity, nuance, trust.

HR intervention should only be necessary when the situation cannot be resolved by a process. If it can be resolved by a process, build the process and let it run.

The failure mode of AI in HR is applying it to human moments.

At a previous organisation we had an involuntary termination. The manager was supposed to have a conversation with the employee before HR could proceed, and treated it like any other offboarding trigger — clinical, process-driven. It backfired badly.

The employee didn't want to follow a process. They wanted to understand what had gone wrong. The manager had run a workflow on a moment that required honesty, empathy and presence. I stepped in, rebuilt trust with the employee, and created a review template for involuntary exits — not to automate the conversation, but to give managers a structure to have it well.

The lesson: AI makes processes better. It cannot make human moments less human. Trying to do so makes them worse.

At an AI-native company, the People function should eat its own cooking.

If you're building AI agents for enterprise clients, your internal HR function should be the proof of concept. Every high-volume, outcome-predictable workflow — onboarding, offboarding, compliance tracking, survey distribution — should be automated. Not because headcount is expensive, but because it demonstrates that the People team actually believes in the product.

The People leader at an AI company should be the first internal customer and the most honest critic of what AI can and cannot do in the context of people.