When the Train Comes Off the Tracks
When performance slips, the instinctive managerial response is tighter control — which systematically destroys the information environment needed to fix the problem it was meant to address.
Performance decline has a predictable secondary effect that is rarely studied as carefully as the decline itself. When organisations begin to underperform, the managerial reflex is almost invariably the same: centralise, tighten control, increase reporting frequency. It looks like corrective action. It functions, in practice, as a second failure — one that systematically destroys the information environment required to diagnose the first.
The structural logic of this spiral is not obscure. But it keeps happening, in sector after sector, because it is emotionally legible in a way that its alternative is not. Control feels like response. Patience, communication, and the willingness to sit with ambiguous data feel, under pressure, like passivity.
The case for crisis control — and its limits
There is a legitimate intellectual basis for authoritarian leadership during acute downturns. A strand of the management literature argues that in genuinely harsh, high-stakes environments, directive control reduces coordination costs, accelerates execution, and eliminates the ambiguity that can paralyse decentralised systems under stress.[1] Some evidence supports this in narrow time windows — particularly in contexts where the workforce has high power-distance orientation, meaning a cultural predisposition to accept hierarchical authority as natural.[2]
But the same literature documents a systematic counter-effect that tends to be under-reported in the short-term readings: emotional exhaustion, accelerated voluntary turnover, and a gradual withdrawal of discretionary effort that the research community has taken to calling cyberloafing — a bland term for the quiet strategic disengagement of people who have concluded that their effort is no longer legibly connected to outcomes.[3] The performance gains, where they exist, are borrowed against future organisational health. The repayment terms are rarely disclosed in advance.
"The Trisolaran authorities had come to understand that the authoritarian regimes built in response to chaotic eras were, in reality, barriers to science."— Liu Cixin, The Three-Body Problem
What Liu Cixin identifies in his fictional civilisation is a structural insight that applies well beyond the interstellar. Regimes built to manage instability tend to calcify into the primary obstacle to the adaptation they were designed to enable. The authoritarian response to crisis is not simply suboptimal — it actively degrades the system's capacity to generate the honest information required to exit the crisis.
A precedent the Romans would recognise
The Romans had a constitutional word for this: dictatura. The office was not a euphemism — it was an explicit, codified suspension of normal governance in response to acute crisis. Its genius, in the early Republic, was its deliberate temporariness: six months, a specific mandate, and then a return to the distributed checks of the Senate. Cincinnatus took the dictatorship, defeated the Aequi, and resigned within fifteen days. The mechanism worked because everyone — including the dictator — understood it as a deviation from the norm, not a replacement for it.
What the late Republic discovered, at considerable cost, was that the emergency instrument corrodes the institutional memory required to operate without it. Each successive crisis — the Social Wars, the proscriptions, the civil conflicts of the first century BCE — made the resort to concentrated power feel more natural, more legible, more efficient than the slower, noisier alternatives. By the time Caesar held the dictatorship in perpetuity, and Augustus reorganised the same power under the more palatable title of princeps, the republican architecture had not been abolished. It had simply become ceremonial. The forms persisted; the epistemic and deliberative substance they once housed had been gradually evacuated.
The organisational parallel is not decorative. Centralisation adopted under genuine crisis pressure has a tendency to outlast the crisis — not through bad faith, but because the conditions it creates (filtered upward communication, atrophied middle-layer initiative, an institutional posture calibrated to compliance rather than diagnosis) make decentralisation feel increasingly risky. The emergency measure becomes the baseline. And the baseline, once established, is very difficult to distinguish from a permanent constitution.
The information problem
The mechanism is worth being precise about, because it is often described in purely motivational terms — morale drops, people check out — when the more serious damage is epistemic.
Under conditions of tight top-down control, the cost of signalling problems upward increases. Team leads and middle managers who accurately report ground-level deterioration face a choice between loyalty to the official narrative and accuracy. When the two conflict, and when the institutional incentives clearly favour the former, the system begins producing filtered data. Leadership calibrates its interventions against a picture of reality that is systematically incomplete — not because anyone is lying in any simple sense, but because the reporting architecture has made accurate bad news structurally expensive to transmit.[4]
This dynamic compounds over time. Impressionistic management of the organisation's public image — emphasising positive signals, deferring acknowledgement of shortfalls, presenting recovery as imminent — creates internal cognitive dissonance for the people who can see the gap between the communicated narrative and the operational reality.[5] That dissonance is not merely uncomfortable. It is corrosive to the psychological contract between the organisation and the people it depends on to execute.
Reporting theatre
One of the most consistent signatures of this failure mode is what might be called reporting theatre: a proliferation of dashboards, synthesis obligations, and time-tracking requirements that function primarily as displays of managerial attentiveness rather than genuine diagnostic tools.
The micromanagement literature is instructive here. Detail-oriented, high-surveillance leadership styles generate measurable costs in employee autonomy, trust, and initiative — even when individual managers adopting these styles believe themselves to be acting responsibly under pressure.[6] The paradox is that the metrics most easily operationalised — hours logged, documents produced, KPIs met on the measurement date — are precisely the metrics most susceptible to local optimisation that is entirely decoupled from underlying performance. The system ends up measuring output appearance while actual output quality continues to decline.
The result is a double bind. Workers are evaluated against metrics that reward gaming. Managers receive data that reflects the gaming. Neither group has reliable access to the signal they actually need.
The AI complication
Into this already-stressed environment, many professional services organisations have introduced AI writing and productivity tools. The adoption dynamics are generative of a distinct additional tension.
Research into AI-assisted writing demonstrates that users simultaneously perceive genuine efficiency gains and experience a significant erosion of authorial ownership — what might be described as a felt loss of creative responsibility for the work produced.[7] The cognitive dissonance is real and documented: people value the tool's practical utility while simultaneously experiencing its use as an attenuation of their professional identity. This tension is most acute when AI generates primary content in domains where craft, argumentation, and original voice are central to the professional's sense of self.
In an organisational context where accountability is already diffuse — where the "who is responsible for this" question is contested across surveillance systems, process layers, and tool outputs — AI adoption without an accompanying conversation about authorial identity and cognitive ownership amplifies the deresponsabilisation dynamic rather than resolving it. The result is a workforce that is simultaneously more productive by certain measures and more alienated from the meaning of that productivity.
The vicious circle
The loop, stated plainly: performance shortfalls produce centralisation. Centralisation signals distrust downward, raising the psychological cost of honest upward communication. Filtered information produces miscalibrated interventions. Miscalibrated interventions fail to address the underlying causes of the shortfall. Morale deteriorates. Performance deteriorates further. The cycle tightens.
The question of whether deresponsabilisation — the diffuse organisational sense that nobody quite owns anything — is a cause or a consequence of this spiral admits no clean answer, because it is both sequentially. It may originate in structural ambiguity: unclear mandates, tool-driven attribution gaps, the genuine collaborative complexity of knowledge work. But once surveillance increases and morale drops, deresponsabilisation becomes a rational adaptation rather than a structural accident. Taking ownership of outcomes you cannot control, in an environment that evaluates you against those outcomes regardless, is not professional commitment — it is a poor bet.
The alternative, which is not complicated
None of what would interrupt this spiral requires the underlying performance problem to be resolved first. It requires only a set of communicative commitments that are cheaper, institutionally, than their absence.
Communicating in comparables — situating current performance against historical data at a scale people can reason about — converts anxiety into comprehension. Distinguishing explicitly between diagnostic monitoring and punitive surveillance is not merely rhetorical: teams know which regime they are in, and conflating the two produces the costs of both without the benefits of either. And protecting high performers from the generalised negative signal — recognising that those most invested in the work are most likely to internalise broad institutional failure as personal indictment — is not a management nicety. It is basic retention arithmetic.
The deeper requirement is a tolerance for honest communication about the gap between the narrative and the data. That tolerance tends to be what crisis conditions erode first. It is also, structurally, the thing most necessary to recover from them. Cincinnatus knew when to hand back the fasces. Most organisations, it turns out, find that part considerably harder.