Eight short lessons on running the people workflows AI has already reached: one requisition traced end to end, one calibration cycle with a real override, and one reduce-or-redeploy decision worked through to the answer. You leave with a one-page decision map naming where each people process pauses, what standard it pauses against, and whose name is on it.
In every other function, a model that gets it wrong tells you. Stock piles up, a margin line moves, a customer complains. In HR the wrong output is a person who never appears: the candidate who was closed without a human reading the file, the employee whose rating was settled by a flag nobody argued with, the team that absorbed a reduction the blended number hid.
So this course is not about the promise of AI in HR, which you already know, or about its risks in the abstract, which every conference covers. It is about the three workflows the function actually runs, worked end to end, with the points marked where a human has to stop and decide something.
You trace one requisition backwards from a wrong rejection to the step where it was really decided. You sit through one calibration cycle where a manager overrides the model and you see what has to be written down for that override to be worth anything three cycles later. You take one reduce-or-redeploy decision and find the thing the headcount model cannot see.
Around those three sit the two jobs that have quietly become HR's: answering the fears people already have about all of this, and teaching the rest of the company to use it on each other.
The deliverable is your HR AI decision map: four people-decisions, and for each one the stop, the standard, and a name.
CHROs and heads of HR who answer for how AI is used on the workforce, whether or not they chose the systems.
Talent, L&D and people-analytics leaders running these tools day to day and deciding where the human step goes.
Founders and COOs whose automation programme is about to land on a function full of people.
Expected time: about 27 minutes across eight short lessons, three to four minutes each.
The judgment stop: a four-part definition of what has to be true for a human step to count as oversight rather than a signature line.
A five-step map of your own hiring loop, and the backward trace that finds where a wrong rejection was actually decided.
The three definitions of a fair model, why they cannot all hold at once, which one employment work tends toward, and how to attach a consequence to the number you pick.
A calibration cycle worked through one real override, with the reason categories that turn a compliance log into a specification for fixing the model.
The reduce-or-redeploy matrix, and the question about load-bearing people that only HR is positioned to ask.
A fear ledger: the affected population's words, sorted by who can answer them, with the changed-design test that separates consultation from a listening tour.
A case-based AI literacy programme for managers, designed assessment first.
Your one-page HR AI decision map, with the weakest row marked for your next supplier conversation.