--- The Death of Process: AI, Organisational Intelligence (or stupidity) and the Future of Work
Most organisations, most of the time, are staffed by ordinary people. Not idiots, but not polymaths either. These institutions depend upon process to function: formalised, repeatable, and predictable sequences of behaviour which, if followed faithfully, produce tolerably reliable outcomes. In such a system, the process is the genius, not the worker. One is hired to fit the groove, not to cut new ones.
This is not a criticism. Process is what allows bureaucracies, firms, and public services to scale. The NHS couldn’t function on the brilliance of individual nurses or administrators alone. Nor could HMRC rely on local judgment calls when assessing tax liabilities. These bodies must coordinate thousands of people with uneven skill levels, incomplete information, and minimal improvisational scope. Process is the cybernetic compromise: a prosthetic intelligence that permits the mediocre to act as if they were better than they are.
But there is a problem. In complex and dynamic environments, fixed processes start to creak. They cannot adapt quickly. They struggle with edge cases. They become brittle, sluggish, or absurd. What begins as a sensible decision-tree often ends as Kafkaesque farce: "Computer says no."
The late cyberneticist W. Ross Ashby famously formulated the Law of Requisite Variety: for a system to effectively control another, it must possess at least as much variety as the system it seeks to govern. In other words, you can't manage complex, highly-contextual situations with simple rules. But that is precisely what most process-driven institutions attempt to do. They simplify and regularise a richly-textured world, not because it's wise, but because it's all they can manage.
This leads to an odd modern duality. On the one hand, we have lumbering procedural leviathans: governments, big corporates, legacy institutions. They crawl forward by the inch, prodded by forms, compliance, and the logic of the lowest common denominator. On the other, we have small, high-functioning firms – consultancies, elite research units, startups – where people talk, reason, improvise. Here process is kept to a minimum, swatted aside when it gets in the way. The organisation runs not on scripts but on minds. When ambiguity arises, it is not escalated to a committee, it is resolved in conversation.
This contrast is not incidental. It is systemic. In smart organisations, top talent vastly reduces the need for mindless process. People can be trusted to do the right thing because they can grasp the whole. They adapt, they course-correct, they bring local intelligence to bear. The result is fluid, context-sensitive action. If a procedure exists, it is provisional, not sacrosanct.
Now enter AI.
We are not yet at the point where large-scale systems can act with contextual intelligence comparable to a sharp human generalist. But we are not far. The next decade promises AIs that can perceive, plan, reason, explain, and adapt at a level sufficient to handle most of the real-world cases that currently clog bureaucracies or consume the time of overqualified analysts.
The result is potentially revolutionary. Consider a large public-facing organisation: the DWP, for instance, or a major utilities provider. Where once it required sprawling call centres and elaborate service manuals to handle customer needs, we could soon see responsive AI agents doing the bulk of the work – triaging, resolving, escalating only the truly complex. Not via static process trees but through adaptive, conversational engagement.
Now imagine that AI not simply replacing average workers, but working alongside skilled human agents: strategists, planners, designers. The AI handles procedural load and problem-solving at scale; the human applies values, ethics, and situational judgment. The result is a high-functioning hybrid.
In effect, this extends the logic of the smart organisation to scale. What was once only possible in boutique firms staffed by high-IQ generalists now becomes viable for large, diffuse entities serving millions. The performance ceiling is raised not by retraining the workforce, but by wrapping it (or replacing most of it?) in a cloud of competent, tireless, adaptive support.
Of course, this raises questions. What becomes of the average worker in such a world? If the mediating layer of process disappears, and with it the need for human cogs in its machinery, what remains for the merely competent to do?
Two futures suggest themselves:
- Displacement with dignity: New forms of work emerge, not in the procedural heartland of institutions, but in care, creativity, education, and human-to-human services.
- Orchestrated redundancy: Workers remain employed, nominally, but become functionally irrelevant. AI handles the real work; humans become face-saving decor, like elevator operators after the buttons were introduced. Gardening leave but at the office.
The smart organisation was once the exception. Soon it may be the norm. Not because everyone gets smarter, but because the system itself does. And in that world, "process" may no longer mean rigidity, but something closer to jazz: structured improvisation, shaped by judgement, and played out in real time.
