Founders keep asking me whether AI is going to run their people function. It isn't. But it will clear a surprising amount off your desk, and the useful question is which parts to hand over and which to keep firmly human.
What AI is genuinely good at now
The repetitive layer of HR — the work that follows rules and repeats every time someone joins, moves or leaves. Drafting job descriptions from a scorecard. First-pass CV screening against defined criteria. Scheduling interviews. Generating offer letters and contracts from templates. Building an onboarding sequence that fires the same way for every hire. Chasing the documents nobody wants to chase.
Done well, this gives a small team back hours a week and takes out the small errors — the missed reference check, the wrong start date, the policy that never got sent — that turn into real problems later.
AI can inform the decisions that compound. It should not be making them.
What stays human
Who you actually hire. What you pay them. The read on someone across a table. The performance conversation that isn't going well. Whether a team is healthy or quietly coming apart. What your culture rewards in practice, not on the wall. These are judgement calls, and they're the ones that determine whether the company scales or stalls. A model can give you inputs. It can't carry the decision, and it can't be accountable for it.
How I set it up
I build the automated layer underneath — the sequences, templates and workflows — so that the human time in your week goes where it earns the most: the hiring calls, the manager coaching, the org design. The point of the tooling isn't to look modern. It's to stop your best people spending their attention on admin.
Where to start this week
List every recurring people task in your business. Mark the ones that are rules-based — same inputs, same output, every time. Automate those first, one at a time. Leave everything that needs a judgement exactly where it is: with a person.