Essay

The company after coordination becomes cheap

AI agents may change not only jobs, but the shape of the company: fewer layers, smaller teams, and denser ownership.

3 min read

Visual thesisA narrower, denser company
  1. 01 Coordinate

    Agents summarize state, dependencies, and routine follow-up.

  2. 02 Escalate

    Humans receive exceptions and decisions, not every event.

  3. 03 Own

    Smaller teams stay close to customers and consequences.

Agents absorb routine coordination; humans concentrate on exceptions, judgment, leadership, and direct ownership of outcomes.

Coordination becoming cheap does not make leadership cheap.

The company is a coordination machine

Meetings move information. Managers translate priorities. Project managers track dependencies. Analysts prepare decisions. Middle layers summarize what the lower layers are doing. These activities are part of the cost of making many people act together.

As companies grow, the org chart becomes a compression algorithm. It reduces communication load, but it also softens bad news, removes context, and creates room for each layer to optimize for its own incentives.

Move status into the background

AI agents can shift coordination from synchronous meetings into asynchronous systems. A maker finishes work; an agent updates project state, checks dependencies, summarizes changes, and prepares the questions that require a person. A status meeting can become a generated report plus twenty minutes on real decisions.

The goal is not maximum information flow. Cheap coordination can create machine bureaucracy—dozens of agents producing plans, alerts, and summaries that consume the day. The goal is the minimum information required for a good decision.

Hire later and higher

Functions that once required employees can now exist in a primitive form before the hire: research, support triage, lead enrichment, security checks, or operational monitoring. This lets a small company stay small while product-market fit remains uncertain.

When hiring finally happens, the role moves upward. Do not hire someone to manually produce reports an agent can assemble. Hire the person who defines what should be measured, interprets anomalies, and changes the system.

Leadership still owns consequences

AI can schedule a difficult conversation, but cannot take moral responsibility for it. It can monitor a metric, but cannot decide what kind of company should exist. Leadership contains coordination, but it is not reducible to coordination.

The future company may be narrower and denser: fewer people performing pure coordination, more people owning outcomes directly. This is especially powerful outside major capital centers, where a company can serve global customers without recreating every department of a much larger corporation.