Peter Drucker on managing knowledge work, 1999
“Still almost entirely unsolved.”
That's his verdict on the one job twentieth-century management never cracked, from Management Challenges for the 21st Century.
The factory got the stopwatch and the conveyor belt: one best way, an output number anyone could read off a shift. You got nothing — and you work inside that unsolved problem all day, every day, grading yourself with the factory's tools because nobody ever built you better ones.
- Cars roll off the line at a fixed rate, and anyone can count them at the end of the shift.
- The task is handed to you before you arrive: one pile, one station, one motion, repeated.
- A supervisor with a stopwatch can watch you, time you, and know exactly how the day went.
Frederick Winslow Taylor, Bethlehem Steel, 1899 — a stopwatch, a pile of pig iron, one best way.
- Your worth is judgement made well under uncertainty that will never fully resolve.
- Nobody hands you the pile. Deciding what today's task even is is itself the first piece of work.
- The most consequential thing you did all week may never appear in any report.
Peter Drucker named this in 1959. Nobody has finished solving it.
This page is long, so here is the part that is actually about you. Everything after it is the evidence: a week nobody can picture, a hundred-year-old model you are still graded by, and one ordinary Tuesday taken apart hour by hour. If you read nothing else, read this.
The naive metric
“What did I ship?”
Counted this way, most weeks look thin. A blocked podcast thread. A grant report paragraph rewritten three times and thrown away twice. A hiring call nobody can confirm was right. A decision that sat in someone else's inbox for a day and a half. None of it looks like output.
What the week actually produced
- A decision made on incomplete information, with the reasoning written down — the one line in the tracker that saves someone re-litigating the same call from nothing next time.
- A blocker named clearly enough to act on — flagged and left alone, rather than chased pointlessly or quietly absorbed as a vague, private worry.
- Tacit knowledge extracted before it was lost — four minutes of someone else's head, turned into two sentences that now exist somewhere other than inside them.
- The paragraph that made someone else's decision nearly automatic — writing done so that a decision two levels up can be made in minutes rather than requiring its own meeting.
- Waiting, correctly — recognising that a call genuinely wasn't yours to make, and that the cost of guessing wrong wasn't yours to carry either.
None of these show up on a ticket count. All five are the actual job.
The reframed end-of-week questions
- Decisions. Which calls did I make on incomplete information this week — and did I write down the one line of reasoning that would save someone re-litigating it later?
- Coordination and structure. Which blockers did I name clearly enough for someone else to act on — and which ones climbed further up, or sat longer, than their actual stakes justified?
- Flow. What did I unblock for someone else this week, even briefly — and would I have noticed it myself if nobody had told me?
- Attention. Where did I protect a real block of thinking, and where did I trade it away to answer something that could genuinely have waited?
- Knowledge. What do I know right now that exists nowhere except in my own head — and what happens to that the week I'm not reachable?
Sitting with those five questions every week, out loud, with someone whose job is to hold the threads with you, is what A Large Company does — support for operators holding more threads than any model tells you how to hold.
The rest of this page is why the reframe is true: the week nobody can picture · the model that doesn't fit · one ordinary Tuesday · the graph beneath it · six lenses
Below is not a company growing. It is one ordinary week inside an organisation of about forty people — headcount fixed, nothing scaling up. It is a failed attempt at picturing that week anyway. Not because the code is wrong, but because the actual thing — every decision, every workstream, every blocker, every quiet piece of uncertainty sitting unresolved in someone's inbox — cannot be pictured at all, by anyone, on any screen. It does not hold still long enough. That is the finding, not a bug.
Start with just the people, before a single workstream or blocker gets added. Forty people produce n(n−1)/2 possible pairwise communication channels — 780 of them, before anyone has done a scrap of work. That is Brooks's arithmetic (more on him later), and it is the floor, not the ceiling. Add the work back in at something like a hundred actions per person per working day, and the week runs to roughly twenty thousand actions, hundreds of decisions made and mostly forgotten within the hour, and blockers opening and resolving continuously, somewhere in the organisation, at every moment you look.
The people. Forty people crackling, projects sitting between whoever currently shares them.
The same week, seen two ways.
The work. The identical week seen from the work's side — projects cycling waiting, active, blocked and done as people migrate between them.
The rest of this site slows this down until it becomes legible — one person, one Tuesday, six ways of seeing.
A man with a stopwatch
In 1899, at the Bethlehem Steel Company in Pennsylvania, an engineer named Frederick Winslow Taylor picked out a labourer to study: a man he called Schmidt in his writing, whose real name was Henry Noll. Noll's job was to load pig iron — rough cast-iron ingots, about ninety-two pounds each — onto rail cars. Taylor timed every component of the task, worked out how much rest a strong man's body needed between exertions, and retrained Noll to work to a schedule dictated from outside rather than from his own sense of his body. Reported output rose from around twelve and a half tons a day to something like forty-seven.
Taylor ran a second study in the same yard: the gangs used one shovel for every material, from four-pound scoops of coal to thirty-pound scoops of ore. He worked out that output peaked around twenty-one or twenty-two pounds a shovelful regardless of material, supplied a range of shovels sized to the job, and the workforce needed to move the same tonnage fell from around six hundred men to about a hundred and forty.
This was scientific management, and for a while it looked like it might travel from the rail yard into the office too. In the 1910s, an American writer named William Henry Leffingwell timed typists, filing clerks and mail-sorters and redesigned desks to shorten the distance a stenographer's hands had to travel — “scientific office management.” Where a clerical job really was a repeated, external, watchable motion, the method worked about as well as it had for shovelling. Where it wasn't, it stalled, because there was nothing left to time.
Drucker names the thing that doesn't fit
By the middle of the twentieth century, Peter Drucker had spent twenty years watching large organisations try to run their offices on factory logic and noticing that the fit was poor. In 1959, in The Landmarks of Tomorrow, he named the mismatch: a new kind of employee, working with ideas, information and judgement rather than hands and materials. He called this person the knowledge worker, and the shift — economies moving from manual to knowledge-based work as their main source of value — the knowledge economy.
In The Effective Executive (1967), he drew the distinction that follows directly from the four invisibles below: efficiency — doing things right, more output per unit of input on a given task — versus effectiveness — doing the right things, choosing which task deserves the effort at all. For Noll, the right thing to do was already settled; only how well he did it was in question. For a knowledge worker, a blisteringly efficient hour spent on the wrong task, while the thing that actually mattered quietly slips its deadline, is a bad week executed extremely well.
He kept returning to the same worry for the rest of his career — The Age of Discontinuity (1969), Post-Capitalist Society (1993) — and by Management Challenges for the 21st Century (1999) had sharpened it into a direct comparison: the great achievement of twentieth-century management had been to raise manual-worker productivity by something like fifty times over. The equivalent task for knowledge work was “still almost entirely unsolved” — and would matter to the wealth of nations at least as much as manual productivity had mattered to the last one.
You cannot time a decision. You cannot hand someone else the choice of task. You cannot inspect a head. The whole discipline Taylor built was an answer to a question knowledge work doesn't ask.
Why the stopwatch worked — and where it stops
Scroll to see both columns →
| Taylor's world (pig iron, 1899) | The work a stopwatch can't see |
|---|---|
| The task is given. A pile of pig iron and a waiting rail car settled the question before Noll arrived. | Nobody hands you the pile. Deciding what today's task even is — which of five open threads needs you first — is itself the first piece of work. |
| The work is external. It happens in the world, in full view, as physical motions a bystander can watch, time and count. | Most of what you do today is reading a message twice, holding two possible phrasings in your head, remembering something someone said three weeks ago. None of it produces a motion an observer could time. |
| The expertise sits with the observer. Taylor didn't need to get inside Noll's head; a stopwatch and a notebook were enough. | Nobody else has the standing knowledge of everything you're carrying, updated in real time. You cannot be handed a schedule the way Noll was, because nobody else could plan your day without first becoming you. |
Four things a stopwatch cannot see
The task is not given. You arrive to several open workstreams and no pile tells you which one needs you first. Deciding what the task is, this hour, is itself work only you can do well.
The work is invisible. Reading an email twice, holding two possible phrasings in your head, remembering an offhand comment from three weeks ago — none of it produces a motion an observer could time. A consultant with a stopwatch would catch typing, a phone call, a walk to the kitchen: the visible five per cent of a day that is nine-tenths thought.
The worker must direct herself. Taylor's whole system worked by moving planning out of the labourer's head and into the engineer's design. Nobody else holds the current state of every thread you're carrying, so nobody else can plan your day for you.
The means of production sit in the worker's head. A factory owns its machines. If you left tomorrow, the actual expertise — who's reliable, what's already been vetoed, how someone likes their drafts flagged — would leave with you. Management no longer owns the means of production it depends on.
Step back from a day like this and a shape appears that the hour-by-hour telling half-obscures: an organisation like this is not really forty job titles arranged on an org chart. It is a graph. Every person is a node, and so is every workstream running through them. The lines between them are dependencies — one person depends on another for a fact only they hold; a third depends on the first for institutional memory only she holds; a producer depends on her, who depends on him; one person's attention is a node that half the organisation depends on at any given moment, which is exactly why it is so often the tightest bottleneck in the whole structure.
Nobody drew this graph. Nobody could, fully — it is too large, too tangled, and it changes shape by the hour. But it is real in the way that matters: it determines, more than any job description does, what gets done and when. This is the same organism as the week at the top of the page, frozen and pruned to a single person's slice — deliberately too simple, so you can finally see it.
Drag or scroll sideways →
Scroll to see the whole table →
| Thinker | Work | Year | Why |
|---|