Work became fragments
Before industrial production, a skilled maker often remained close to the whole arc of the work. A craftsperson could understand the material, make consequential choices, see the result in use, and carry a reputation attached to it. This world was never universally free or prosperous, and craft itself was often protected by exclusion. But it preserved a relationship that modern organizations frequently weaken: the relationship between intention, action, consequence and authorship.
Industrial organization created extraordinary scale by dividing that arc into repeatable parts. Specialization made complex products possible and raised material productivity. It also established a powerful institutional habit: break work into positions, move partial outputs between them, and coordinate the whole through hierarchy, process and management.
The office inherited this logic. The conveyor belt became the ticket queue. The workstation became the job description. Physical components became requirements, presentations, approvals, code changes and reports moving across functional boundaries.
Each participant can perform a local responsibility correctly while the complete outcome still fails. Context fades at every handoff. Decisions become detached from consequences. When success arrives, authorship is diffuse; when failure arrives, responsibility can dissolve into evidence that every team followed its process.
When nobody can see or influence the whole, ownership becomes a procedural claim rather than a lived relationship with the result.
This is not a moral failure by workers or managers. It is an architectural consequence of how organizations learned to scale under limited communication and cognitive capacity.
Why removing hierarchy was not enough
Many organizational movements recognized the damage caused by fragmentation. Cross-functional teams, agile methods, flatter structures and distributed authority all tried to return decisions to people closer to the work.
They produced important improvements, but they could not repeal the underlying coordination problem. A person still had limited time, attention and specialist knowledge. When formal hierarchy disappeared without a corresponding increase in individual capability, coordination often returned as meetings, negotiation, informal influence and continuous role clarification.
The lesson is not that hierarchy is inherently good or that self-management is impossible. It is that organizational form cannot be separated from material capability. Removing a structure does not remove the work that structure was performing. If people still need many other people to complete one causal chain, some mechanism must coordinate their interdependence.
What intelligence changes
Intelligent systems introduce a different possibility. They can help one person or a small team research across domains, examine large bodies of material, produce and test alternatives, operate tools, monitor systems and carry context between stages of work.
The important change is not simply that one person can complete more tasks. It is that more of the execution chain can remain close to the people who hold the original intent.
Specialists do not become unnecessary. Collective work does not disappear. Some outcomes require deep expertise, institutional legitimacy, independent review and many accountable participants. But collaboration can increasingly happen around complete outcomes rather than around the transfer of fragments.
Used well, intelligent capacity can reduce the distance between a question and its consequence. It can allow a person to remain involved from definition through construction, inspection and revision. It can make the organization less dependent on humans acting as information routers between disconnected systems.
Used badly, the same technology can intensify fragmentation. It can produce more output without understanding, hide consequential decisions behind automation, and concentrate control in systems that nobody affected by them can inspect. Greater capacity does not automatically produce better institutions.
A modern form of craft
The return of craft is not a return to a pre-industrial economy. It is not nostalgia for the solitary artisan, nor a claim that one generalist should replace every profession.
Modern craftsmanship is a relationship between a person, a complete outcome and the systems that help create it. It has five qualities.
Intent means understanding why the work should exist and whose life it may affect.
Wholeness means staying connected to the causal chain rather than optimizing one isolated step.
Taste means exercising standards that cannot be reduced to speed, volume or formal compliance.
Responsibility means accepting that authority over an outcome carries obligations when the outcome succeeds and when it fails.
Signature means that authorship remains visible. Not every contribution needs an individual name, but every consequential result needs people who can truthfully say, “We understand this, we chose it, and we stand behind it.”
Intelligent systems can extend the maker’s reach, but they cannot supply the moral basis for the work. They can search, compose, simulate and execute. They cannot decide which future deserves to be built or make responsibility disappear when the result enters society.
Responsibility without the heroic individual
Restoring authorship must not become a new mythology of the sovereign individual. Important work is shaped by communities, inherited knowledge, public infrastructure and the labor of people who may never be visible in the final signature.
End-to-end ownership should therefore mean clear stewardship, not unchecked control. The responsible maker remains accountable to peers, affected communities, law, evidence and independent challenge. Boundaries, review and escalation are not obstacles to craft; they are part of practicing it in a connected society.
The goal is not to collapse collective responsibility into one heroic person. It is to prevent responsibility from disappearing into the organization. A healthy institution can preserve both: identifiable human stewardship and meaningful collective governance.
A social choice
Technology does not determine which organizational future we receive. Intelligent capacity can enlarge human agency, or it can make people more dependent on distant platforms and opaque decisions. It can broaden access to institutional capability, or concentrate that capability among organizations already powerful enough to own the infrastructure and data.
The direction depends on institutional choices.
Who can use these systems? Who can inspect them? Who retains the knowledge produced through their work? Who receives the economic gains? Who can refuse their decisions? Who remains answerable when automated execution causes harm?
Dark Offices begins from a commitment: expanded intelligence should help more people become capable authors of meaningful outcomes. It should reduce unnecessary dependence on place, hierarchy and organizational size while making responsibility more visible, not less.
That is larger than a productivity programme. It is an attempt to support a different social settlement around capability and work.
Questions that remain open
We do not yet know the stable organizational forms that will emerge. We should be suspicious of anyone who claims to know them with certainty.
How much execution can remain close to one accountable maker before independent review becomes too weak? How do we preserve the benefits of specialization without returning to fragmented ownership? How should organizations distribute the gains when a small team can produce what once required hundreds of people? What forms of education prepare people for judgment and authorship rather than fixed occupational slots?
These are not objections to the transition. They are part of the work.
The promise is not that intelligent systems will make organizations effortless. The promise is that we may no longer need to organize every person as a replaceable step in somebody else’s process.
Work can become whole again. Whether it does is a choice we still have to make.
Intellectual foundations
This essay develops an argument informed by Ronald Coase’s work on the boundaries of the firm, Alfred Chandler’s account of managerial coordination, Herbert Simon’s theory of bounded rationality, and Daron Acemoglu and Simon Johnson’s analysis of how institutions shape the distribution of technological progress. These foundations frame the questions; they do not establish that any particular organizational future is inevitable.