The point of view
Waste you couldn't see. Waste you couldn't touch.
Knowledge work has always resisted the discipline that transformed the factory floor. Inference changes what can be made explicit — and that changes what counts as waste. Here is the argument, laid out plainly.
The old problem
Work done well once can't be repeated.
Every knowledge organization knows this failure. The expertise that drives excellent work lives in people's heads — experience-shaped, hard to articulate, transferable only by apprenticeship. Documentation exists, but it describes the ideal process rather than the practiced one, and the gap widens until the document is fiction.
The traditional responses — hire better, train harder, inspect more — don't scale, and your system already knows it. What Nonaka and Takeuchi called externalization, turning tacit knowledge into explicit knowledge, was always the right move. It was just too labor-intensive to justify for most commercial work. That is the constraint that moved.
What changed
An execution problem became an engineering problem.
A model in a well-built harness, guided by an expert's method made explicit, can now carry real commercial work — retrieval, synthesis, drafting, consistency-checking — while people carry what people are for: judgment, exceptions, relationships, taste. Neither is better; each is suited. The design question is the old lean question in new clothes: right work, right hands, quality at the step.
Two consequences follow, and they are the whole argument. Work that could never be observed becomes observable — the redesign itself is the instrument. And fixes that could never be justified become economical, because the cost of making a method explicit fell by an order the spreadsheet can see. New waste is visible. New countermeasures are viable. The benchmark for well-run knowledge work moved.
"This capability is new to everyone. When a genuinely new discipline appears, the strongest operating systems have always absorbed it the same way: bring in a master, learn by doing, make it your own."
The Five New Wastes
The taxonomy, in full.
Your tradition names seven wastes of the factory floor. These are the five we walk for in knowledge work — each one either invisible until a process is redesigned AI-native, or visible but never worth fixing until now.
Unpackaged Context
Every process we walk has a step that only works because someone remembers. A person muddles through it; a machine fails it outright. The cost was always there — rework, re-explanation, answers that sound right and aren’t — but it never had a line item until work started moving between hands.
The question we ask on the walk: “Which step of this process only works because someone remembers?”
Fixed, it looks like: context packaged at the point of work — sources, constraints, and prior decisions travel with the step instead of living in someone’s head.
Trapped Expertise
Your best work happened once; the method lives in one expert’s habits. Apprenticeship was the only transfer mechanism, so the organization re-solves solved problems and the expert becomes the bottleneck. Externalization was always the answer — it was simply never economical for commercial work.
The question we ask on the walk: “Whose judgment makes this good, and could anyone else run it their way?”
Fixed, it looks like: the expert’s method made explicit enough to run — by a colleague, or by a machine under the expert’s standard.
Misassigned Work
People doing retrieval, synthesis, and consistency-checking; machines handed judgment, exceptions, and ethics. Both sides lose — cost on one, quality on the other. This waste literally could not exist before there were two kinds of hands to assign work to. It is the newest of the five, and the most common.
The question we ask on the walk: “Where are your people gathering and reconciling instead of deciding?”
Fixed, it looks like: an explicit assignment for every step — human or machine — with the escalation path named, so nothing lives in the ambiguity between.
Late Inspection
Your system won this fight on the factory floor decades ago: quality built into the step beats inspection at the end. Machine-assisted work reopened it — output that reads clean can still be wrong, and the defect surfaces far from where it was made. The old countermeasure applies; it just has to be re-installed on new work.
The question we ask on the walk: “Where is this checked once at the end, instead of at the step?”
Fixed, it looks like: acceptance conditions at each step — “done” as a condition met, not a feeling — so defects surface where fixing them is cheap.
Attention Overload
Knowledge work’s binding constraint is attention, not throughput. Processes that spend it carelessly degrade quietly, and nobody traces the output back to the overloaded step. The factory learned to design work to fit the worker; knowledge work mostly never did — because the load was invisible. In a redesign, it isn’t.
The question we ask on the walk: “Which handoff quietly taxes the operator’s attention?”
Fixed, it looks like: work designed so the next correct action is obvious — the machine carries the recall, the person keeps the judgment, and quality stops leaking through fatigue.
Why the walk finds it
Your discipline already works here. That is the point.
Define value. Map the stream. Make it flow. Let it pull. Improve the standard. The five principles your system runs on apply to knowledge work exactly as written — what was missing was the ability to see the stream and afford the fix. The walk is the same walk; the five wastes are what it finds on this terrain. And because the redesign can now be prototyped while it is being designed, the standard that comes out of the event is not a binder. It runs.
See how the event runs →Argued in public
This view is made in the open.
Presented at university-hosted, co-sponsored sessions and industry stages — the same argument you are reading here, defended in front of the people who run the work. A point of view earns its name by being tested aloud.
Walk one process against this list.
Tell us where the work slows. We'll show you what the event would run against — before you commit to anything.