The CARE Loop in Practice: The Schedule Was Legal, Efficient, and Wrong
The AI shift-allocation system Audrey’s company installed did everything the procurement deck promised. Across two Kuala Lumpur warehouses, overtime costs fell, coverage gaps closed, and the roster that once took planners two days assembled itself in minutes. Six months in, Audrey — who leads HR for the logistics firm — was looking at a different dashboard: resignations up in exactly the two facilities where everything was supposedly working, exit interviews full of the word “schedule,” and a supervisor who finally said the quiet part in a corridor: “The roster broke the carpools, and we all pretended not to notice.”
Nothing about the system was illegal. Nothing about it was even unreasonable, shift by shift. What it had done was optimize away a layer of life the data never carried: the carpool clusters that got single mothers home safely from the night shift, the swap culture that let a packer attend Friday prayers, the stable-shift veterans whose predictable weeks were the quiet reason their households functioned. The system saw workers as schedulable units of one. The warehouse floor is not made of units of one.
Chapter 9 of the book offers teams a practice for exactly this gap — the CARE Loop, four movements, run as a standing weekly discipline rather than a crisis response. Audrey’s team now runs it every Thursday, thirty minutes, one AI-touched process at a time. Their first loop, on the scheduler:
Context — map the whole system
Not “warehouse staff” but the actual worlds colliding in one roster: the four carpool clusters and who anchors them; the workers for whom Friday midday is non-negotiable; the two hundred employees on shared or family phones who never see schedule changes pushed after 9 p.m.; the planners themselves, who had quietly become the system’s apologists because it had also eliminated their most hated task.
Acknowledge — name impacts honestly, in the room
This was the hard movement, because it required saying findings aloud in a team setting with supervisors present — the book’s point being that invisible costs stay invisible when only leadership hears them. The corridor sentence about carpools was said again, on the record. So was the fact that attrition among night-shift women was triple the site average, a number that had technically been available to everyone and looked at by no one.
Respond — act at two scales
Immediate, within the fortnight: carpool clusters entered the system as scheduling constraints; the shift-swap window the old paper process allowed was restored inside the new one; late-night schedule pushes ended. Systemic, over the quarter: worker preference inputs were added to the model — a field the vendor supported all along but no one had turned on — and a fairness review of the roster now runs quarterly, chaired by operations rather than HR, which Audrey insisted on: “Care that lives only in HR is a complaint channel. Care that lives in operations is a design input.”
Evaluate — metrics and stories together
The loop closes with both kinds of evidence: attrition and absence numbers on one side, and a monthly Reflection Hour on the other, where planners and floor staff review what the roster did to actual weeks. The stories keep finding what the metrics blur — the most recent Reflection Hour caught the scheduler quietly assigning the newest hires every unpopular shift, a pattern perfectly rational to the optimizer and perfectly corrosive to a shift floor’s sense of fairness.
The scheduler stayed. That’s worth underlining: the loop did not un-buy the software, and the efficiency gains did not evaporate when the constraints went in — they shrank by an amount the CFO described as “noise.” What changed is that the system now optimizes inside boundaries a human team consciously chose, reviews, and owns. The schedule is still made by a machine. What it’s made of — that’s back in human hands.
Audrey is a composite — drawn from conversations with HR and operations leaders deploying workforce AI in Southeast Asia and elsewhere, with details changed. The corridor sentence has been said, in some form, in every industry.
Try it yourself
Pick one AI-driven process that touches people your team is responsible for. Run the four movements in thirty minutes — map the actual worlds colliding in it, name the impacts out loud with the affected people’s supervisors in the room, act at two scales, and schedule the evaluation before the meeting ends. Then make it weekly. Care, the book argues, is not a sentiment; it’s a cadence.
Comments