Late one night in 2002, in a conference room at a Toyota factory, a Seattle doctor named Robert Mecklenburg was drawing a diagram. A sensei, a master teacher of Toyota’s production system, had asked how patients moved through care at Virginia Mason Medical Center, and Mecklenburg was sketching the path.
The sensei asked about one of the boxes. That was the waiting room, Mecklenburg explained. Patients waited there for the doctor, for 45 minutes or so, even with a scheduled appointment. The hospital had at least 40 of these rooms, each 25 by 30 feet. Then the interpreter turned to Mecklenburg: “Sensei wishes to know, aren’t you ashamed?”
The waits, the fish tanks, the coffee machines and the staff juggling appointments, Mecklenburg later told PBS, “were all unnecessary costs which are ultimately paid by that patient.” Ten years on, one of the hospital’s clinics had no waiting room at all. The case has the rest.
Seven names for waste
Waiting, hauling, extra stock and rework cost money, and customers rarely gain from them.
According to the Asian Productivity Organization’s glossary, Taiichi Ohno, a Toyota engineer, sorted waste, muda in Japanese, into seven kinds:
- Overproduction: making more than is needed or ordered.
- Waiting: for a machine, for materials, for a person who is late.
- Transporting: moving things from place to place.
- Processing: steps built into the process or design that need not be there.
- Inventory: stock and work in progress that tie up money and space.
- Motion: people or equipment moving without adding value.
- Defects and rework: fixing bad work and handling complaints.
The labels travel. In a hospital, the Virginia Mason Institute suggests, motion is a nurse leaving to hunt for supplies that could have been kept at hand, a defect can start with a doctor’s illegible handwriting, and processing waste is a report nobody needs anymore.
Movement is not work
Ohno’s own writing is less about lists than about learning to see. Ohno told supervisors to train themselves “to distinguish between movement and work,” which Ohno called cultivating “an ability to find waste.”
Ohno was hardest on overproduction. Make only 100 items if you need 100, Ohno wrote, “even if you could make 110 of them before quitting time,” because making the 110 “would end up being more expensive.” People, Ohno noted, will protest that it is wasteful not to keep producing while materials are still on hand.
At Virginia Mason, the hospital administrator, Charleen Tachibana, told PBS that nurses had been spending too much of their time collecting supplies and drugs. The hospital moved supplies out of a central storeroom and into patient rooms. Through a variety of changes, by Tachibana’s account, nurses’ time in direct contact with patients rose from 35 percent to upwards of 90 percent.
What the evidence shows
By 2012, PBS reported, a number of hospitals were experimenting with Toyota’s principles, often under the name Lean. The research on Lean in health care is less tidy. A 2016 systematic review by John Moraros and colleagues included 22 studies of Lean in health care that passed its quality check, and separately analyzed public data from the Saskatchewan Health Quality Council and the Saskatchewan Union of Nurses in Canada. None of the 22 studies directly measured patient satisfaction or actual financial costs; those judgments drew on the provincial data. Weighing it all, the authors concluded that patient satisfaction and health outcomes showed no significant improvement, that the findings on costs and worker satisfaction were unfavorable, and that gains in things like patient flow and safety were possible but inconsistent. “The evidence to date,” the authors wrote, “simply does not support” the claim that Lean leads to quality improvements in health care.
That doesn’t make waste imaginary. A 45-minute wait for a booked appointment is a cost, whoever pays it. But spotting waste is not the same as removing it well, and some of what looks like waste may be slack that keeps a system safe on a bad day.
Imagine an office team that runs a “waste walk,” finds that people spend hours each week copying numbers from one system into another, and cuts the step. If nobody checks what that step was quietly catching, the first wrong invoice may be how they find out.
Waste in an agent’s context
An AI agent’s working memory, its context window, is a finite space, and Claude Code’s documentation lists ways people fill it with things that add nothing. Ask Claude to “investigate” without limits and it “reads hundreds of files, filling the context.” Correct it again and again, and the context fills with failed approaches. Mix unrelated tasks in one session and the context is “full of irrelevant information.”
The documentation doesn’t call these waste. But the first looks a lot like motion, the second like rework, and the third like inventory that is stacked up and in the way. Its fixes echo Ohno’s: scope the search, start over with a better instruction after two failed corrections, and clear the context between unrelated tasks.
In Seattle, the waste was in plain sight: 40 rooms, built for waiting. It took a stranger’s question to make it look like waste.