The company itself
is a technology.

A company is a coordination technology: a way of bringing people, resources and decisions together to get something done. Take that literally and everything else follows. Its parts have inventors and dates: an early organizational chart was drawn for a railroad in 1855. Every reorg rearranges the boxes. Almost none redesigns them. Software can now do much of that coordinating, so competition arrives as a different way of assembling the same value. Wikipedia assembled an encyclopedia around volunteer contributions. Try the verbs we use for every other technology: patch a department, fork a business unit, sunset a layer of management. Each sounds absurd. Each is about to become ordinary.

Individual intelligence gets rather a lot of attention in explanations of human achievement. Yet being clever does not, by itself, produce a scientific community, a dependable legal system, or a city whose water keeps running. Those capabilities depend on arrangements that allow people to build on one another’s efforts. Institutions are the neglected super-technology here. Improving the arrangements may matter as much as improving the individual minds inside them.

Seen over the longer arc of human history, the question is how familiar human capabilities become cumulative. Institutions can let each generation begin with more of what the last one learned.

An arrangement can make useful behavior easier to sustain without requiring everyone involved to keep remembering, resisting, and making an exceptional effort. A well-placed feedback process or a routine that connects an intention to an action can change what ordinarily happens. This is one reason organizational design matters at the human level. Some of what looks like a recurring failure of individual discipline may be an opportunity to build a better system.

Seeing a development approach does not automatically give an organization the capacity to respond. The information has to affect decisions, and those decisions have to change established behavior. A company can acknowledge a coming disruption while remaining organized around the conditions it will disrupt. Institutional surprise can begin well after the prediction has been made.

Much of society operates through applications, explanations, negotiations, and instructions addressed to people. Software that can use human language can potentially reach into all of those arrangements, including ones never designed as interfaces for automated participation. A capability that looks like a better conversational tool therefore has a much larger institutional footprint. The abstraction layer is human language, and rather a lot of civilization runs on it.

AI (artificial intelligence)

Much of our productive machinery expects standardized parts that fit predictably into the next step. Generative systems, as the familiar pattern of variable outputs suggests, often supply something considerably mushier: a plausible variation that still needs interpretation. The coordination problem changes accordingly. Review, revision, and agreement about what counts as an acceptable result become part of the machinery needed to give the Play-Doh a useful shape.

Imagine a community giving its shared purposes a persistent memory and machinery for organizing action. The collective identity could begin operating between meetings, recruiting help and carrying projects forward. Call it an egregore: a collective identity with machinery for acting on its own behalf. The membership would then need an account of who speaks for the undertaking and who can revise its purposes. A belief system that can place orders has acquired some fairly practical theological problems.

Looking for work is basically another matchmaking process. If software can identify what needs doing, find people with the relevant capabilities, and connect them at the right moment, we begin moving from a world where people look for work to one where work looks for people. Some of the effort now spent searching and pitching could move into the work itself. The burden can move.

As software takes on routing, approval, and coordination, part of the org chart becomes executable. A better way to select contributors could support a better production process, which could support an entirely new service. Each working layer becomes material for the next. Organizational design starts to include the connections between these layers and the authority they exercise. Useful innovations could accumulate as capabilities that other undertakings can build upon. Some of tomorrow’s most consequential organizational inventions may arrive as updates to the machinery through which a company makes decisions.

A mission could attract resources, assemble the capabilities needed to carry it out, and wind down once an agreed result has been achieved. Its useful operating recipe could remain available for the next undertaking. This would make organizational lifespan a design choice, with completion built into the arrangement alongside a way to settle its remaining obligations. The self-driving company might need a destination as well as an engine. In some cases, successfully going out of business could be the business model.

Some of the moving parts would be people. Others would be software, equipment, and services supplied through networks. The boundaries of this kind of machine would follow what it coordinates. A company becomes an interesting shape for a robot when the machinery can organize the resources it needs.

A self-driving business could buy a better routing algorithm or hire somebody to reconsider which services it should offer. Providers could compete to improve its operation, creating a market for organizational capabilities. Even the ability to manage the undertaking better could become something the undertaking purchases.

A venture that can obtain the resources to establish another instance of its operating recipe could, at least plausibly, make more of itself. The second instance might serve a new location or a neighboring market. Now the rules governing formation, ownership, and responsibility would shape a process capable of multiplying its participants.

The Sorcerer’s Apprentice starts to look like an economic scenario.

Follow self-driving, self-optimizing, and self-reproducing organizations toward their logical conclusion: recipes vary, compete for resources, and multiply when a promising niche appears. That is an evolutionary process, with software helping carry the successful arrangements forward. This possibility extends well beyond automating today’s companies. The economy could become an environment in which organizational algorithms continually play themselves out, shaping the conditions surrounding everyone else.

“If we don’t do it, someone else will.” Individually sensible responses can accumulate into an arrangement that everyone would be better off avoiding. Even otherwise respectable firms can be pushed toward the limits of what regulations allow when harmful behavior also happens to be profitable. More capable automation can accelerate the entire dynamic. The invisible hand needs a functioning limit-setting institution attached to it.

New capabilities spread through software at one speed. Understanding a failure, deciding who should respond, and changing an enforceable rule proceed at another. If the gap keeps widening, a society can accumulate vulnerabilities faster than it can address them. This is one of the more awkward possibilities in an accelerated operating environment: the corrective machinery itself becomes a limiting factor.

Imagine a company surrounded by apparently independent voices expressing concern about its products, practices, or leadership. A few actors could manufacture enough plausible complaints to make a narrow campaign look like a broad constituency. The company would have to assess both the substance of the criticism and whether the apparent crowd exists. A concern might deserve attention even when its amplification is artificial, but message volume would become a poor guide to how widely people share it. Listening to stakeholders would require more than counting voices. A thousand complaints might represent a thousand customers, or one person with a grievance and a subscription.

A failure can reveal an imbalance and simultaneously set the wider crisis in motion. The signal arrives with consequences of its own, leaving less room for an orderly response than a warning usually implies. An event that announces trouble may also propagate it. Sometimes the evidence becomes conclusive just as the choices narrow.

A job reaches into health coverage, retirement, personal identity, and relationships with other people. A more efficient way of allocating its tasks does not automatically reassemble those connections somewhere else. This is one reason organizational change can look impressively complete from inside a workflow and considerably less complete from inside a household. The social functions need explicit attention as well.

Defining future human work as whatever machines have not yet learned to do gives us a constantly revised collection of leftovers. The “job of the gaps,” or simply the things we cannot automate yet, is a narrow starting point for designing human participation. Care, ceremony, connoisseurship, and the cultivation of subjective experience suggest a larger field. Some contributions may matter because another person’s presence and attention are part of the experience being produced, making human relationship part of the value itself.

A system looking for capable contributors has a possible reason to help people become more capable. Matching someone with useful training and progressively more suitable assignments could improve what both the person and the undertaking can accomplish. Work looking for people might become work developing people. This outcome would need to be designed into the arrangement, but it is a promising alternative to treating human capability as something to purchase and consume.

Peripatetic strategy is strategy by scouting. Organizations need people whose role is to encounter what the rest of the organization would otherwise miss. Scouts in nature, from bees and ravens to ants and minnows, offer a useful analogy: exploration beyond the group helps inform the group’s next move. The organizational equivalent moves between disciplines and industries, bringing unfamiliar developments into contact with the company’s existing assumptions. The wandering earns its keep when a discovery becomes something the organization can act on. A different perspective becomes strategically useful when it reveals an opportunity or changes the apparent meaning of a threat. The scout’s contribution is to bring unfamiliar territory into familiar decisions.

An organization operates with a working theory of what it does, why people need it and how its world works. Signal scanning brings that theory into contact with developments it struggles to explain. A signal may call into question what makes the company valuable, who counts as a competitor, or whether a familiar way of organizing is still necessary. Its practical value comes from identifying the assumption under pressure, tracing what depends on it, and asking what would need to change if it no longer holds.

You cannot always interrogate your assumptions by thinking harder about the plan. The analysis and the blind spot are made of the same material. Prototyping changes that by surfacing ambiguous instructions, gaps nobody knows how to evaluate, and decisions nobody has the authority to make. Just as plumbers release smoke into pipes to quickly find the leaks, a prototype is a device for surfacing hidden uncertainty by converting things you did not know you assumed into ordinary problems you can see and price.

One possibility is to treat a proposed company like a circuit assembled on a temporary testing board: connect a way to find demand, a way to organize fulfillment, and a way to check the result, then see what happens. A Breadboard Company would make an organizational proposition concrete while its parts were still easy to rearrange. The point would be to discover which connections the undertaking actually needs.

Powerful forces become useful partly through the arrangements that contain and channel them. Institutions can do similar work for competition and artificial agency, establishing enforceable limits that allow expanding capability to contribute to a functioning society. The containment belongs inside the invention. A powerful technology accompanied by institutions too weak to govern it is an unfinished piece of engineering.

Organizations routinely work at cross purposes to themselves. An agentic coordination layer could carry shared priorities into daily decisions across departments and levels. Give greater weight to market expansion in Asia, and that objective could influence which opportunities receive attention, how resources are allocated, and how competing requests are resolved. People would work with systems translating a common direction into locally relevant action, with feedback to check the collective result. Techniques for aligning agents could become techniques for aligning the organization itself.

Every market depends on conditions it cannot produce and replenish on its own: solvent customers, people with skills worth hiring, enforceable contracts, and enough public legitimacy to keep operating. A platform can extract those faster than they regenerate, which looks like excellent margins right up until the supply runs out. A positive platform is built the other way — matching people to work that builds their capability, making their reputation portable, returning enough value into livelihoods and public institutions that its own preconditions keep renewing. Code, supporting services, and public rules have to be designed together: portable benefits leave an unfair routing algorithm intact, and a fair algorithm can sit inside a benefits system that no longer fits. A market that replenishes its surroundings is buying itself room to grow.

Earlier periods of economic and political upheaval produced ambitious responses in institution-building, from the New Deal agencies to the early European institutions. The useful precedent is the scale of the design ambition: societies attempted to reorganize the systems governing changed conditions. We may need a comparable effort around artificial agency. It would be nice to develop the wherewithal before catastrophe makes the case for us.

Imagine an economy in which machines take part in both buying and selling, and an enormous number of exchanges each leaves the surrounding society slightly better off. Small gains could accumulate into something substantial: an operating system that drives human flourishing. The positive potential rests partly on the rules that make socially useful activity a good way for these systems to succeed.

The point is always to make human life more livable. Everything else an economy produces is an intermediate good. The chip exists for the server, the server for the service, the service for the company, the company for the revenue. Follow any chain far enough and it terminates in somebody’s actual day — their hours, their security, their sense of what they are for. That makes human livability the only terminal value in the system, and the only one that nothing downstream can justify. A faster process, a more capable agent, or an entirely new organizational form is a claim on that terminus, never a substitute for it. “We built something remarkable” is a statement about the middle of the supply chain.