The planted saboteur was redundant. Management had already installed everything on his list.
In 1944 the OSS sabotage manual told agents how to slow an enemy organization without getting caught: insist on doing everything through channels, refer every decision to a committee, haggle over precise wording, demand written orders. Most people who read it recognize their own workplace. The natural conclusion is that their organization is either badly run or quietly undermined by someone acting in bad faith. Usually neither is true, and the real reason is more troubling.
Picture two factories. In Factory A, an enemy agent has been planted to slow everything down. He insists on triple-checking every order, routes decisions through committees, demands written approval for routine tasks, and questions the wording of every memo. In Factory B there is no agent. But management, terrified of liability and obsessed with legibility, has independently instituted triple-checking, committee routing, written approvals and precise wording standards. Both factories produce at the same dismal rate. The saboteur in Factory A is superfluous. Factory B is sabotaging itself with more enthusiasm than any enemy could muster.
Why intent doesn’t matter
Factory B’s managers aren’t fools. Each of their rules answers a real need, and every one of those needs is control. Routing through channels makes activity visible to whoever sits above the system. Committees distribute accountability, which protects individuals at the cost of speed. The written record is the medium of legibility and the weapon of retrospective blame. And whoever controls the wording controls the outcome.
So the problem isn’t bad actors exploiting a good system. The system is optimized for one goal, control, that is structurally opposed to another, output, and any system sufficiently optimized for control will produce the same behavior patterns as deliberate sabotage, because control and sabotage share the same operational grammar. The agent in Factory A chose these tactics because they slow things down. Factory B’s managers chose them because they make work legible, and the slowdown came free.
The convergence then accelerates. The people most capable of producing output are also the most frustrated by the process, so they leave first. Through this Dead Sea Effect, the residue who are most comfortable with process-as-work entrench and write the next round of rules. Over time Factory B’s workforce evolves toward people for whom the sabotage manual describes not dysfunction but normalcy.
Why the best managers can’t see it
The twist is that the better an organization gets at control, the more perfectly it replicates sabotage behavior, and the less able it is to recognize what is happening — because recognizing it would require questioning the very competencies that define institutional excellence. The traits that get a manager promoted in Factory B, being process-oriented, documentation-heavy and accountability-focused, are the same traits that make the manual’s instructions seem like common sense. The most locked-in person is the one who reads the manual and thinks, “yes, this is just good management.” They are not wrong about what the manual describes. They are wrong about what it produces.
That is why the usual fixes fail. Burning down the bureaucracy throws away the coordination that process legitimately provides; Factory B does need to know what it ordered. Better process just builds a more refined instrument of control. The real diagnosis is that convergent sabotage is not a process problem but an optimization target problem. Fields selected against failure rather than for control converge too, on a different set of habits that catch the operator’s mistakes instead of assigning them. An organization optimized for output has to tolerate the illegibility, disorder and loss of control that output requires: managers who cannot track everything, committees that do not approve everything, and work that happens in ways no documentation captures.
Few organizations ever ask whether they are optimizing for control or for output, because the honest answer would threaten every career built on process expertise. If Factory A’s saboteur applied for a job at Factory B, would anyone there recognize him? More likely they would promote him.
Threads to Pull
- Jerry Z. Muller, The Tyranny of Metrics: metric fixation produces the manual’s pathologies across healthcare, education, policing, and the military.
- David Graeber, Bullshit Jobs: a significant fraction of workers judge their own jobs pointless, which is what living inside the manual feels like.
- Stafford Beer’s Viable System Model and Project Cybersyn: an attempt to distribute control recursively so an organization keeps coordination without converging on sabotage.