Branch

Politics of Observability

By dan

Human dan

This conversation is a discussion of this new paper.

Abstract

Coase explained the firm as the answer to transaction costs. Work stays inside when specifying, verifying and directing outsiders costs more than hiring employees. This note argues that AI erodes the assumptions behind that answer, rather than merely lowering the costs. Engineered systems known as harnesses largely complete contracts in real time, make external work more observable than employment, and carry context and control across the firm’s edge. As a result, the boundary that predicts where context sits, who controls the work, and where knowledge accumulates becomes the instrumented system, not the legal entity. Legal services, the profession organized most completely around Coasean frictions, is the worked example. Three tests decide the order in which work migrates. Three un-Coasean consequences follow. The boundary becomes something providers compete to move, knowledge pools with them, and incentives invert toward legal dispute prevention. Four conditions identify which other sectors restructure. The argument is predictive because the external-harness model is still early, and the pace is set by organizational inertia rather than the technology itself.

Three Erosions

1. Incomplete Contract Now Completes Itself

Contracts were incomplete when specifying every contingency was impossible, and the firm existed in part because a contract for "handle whatever arises" could not be written. Instead a general counsel was employed. Within an engineered AI system, what practitioners now call a harness covers context retrieval, encoded policies, deterministic rules, logging, and escalation gates. It thereby applies the client's encoded policies, sometimes in a probabilistic (so non-deterministic way) to cases nobody foresaw, escalates what is genuinely novel, and records every step and the rationale. In functional terms the harness acts as a contract that largely completes itself in real time. The classic reason to employ rather than contract weakens in proportion to what the harness can carry. Residual incompleteness remains for genuine novelty; the harness does not eliminate the need for human authority at the edge.

2. Observability Inverts

Williamson's hazard, and Alchian and Demsetz's account of the firm as a monitoring device, assumed that what is outside is opaque, and the employee can be monitored. The AI inversion is that a harness-delivered, external service is more observable than an employee: every retrieval, every check, every draft, every sign-off is logged and inspectable, while your own employee's reasoning cannot be audited at all. The make-or-buy hazard was moral hazard in the market from unsupervised external providers. In the AI era it is becoming relatively opaque internally. Buying becomes the more transparent option, given the
right interfaces.

3. Boundary

The boundary that matters stops being the firm. Coase's question was which transactions come inside the legal entity. With AI the question now is which decisions come inside the
instrumented, policy-governed, logged system, and that boundary cuts across entities. A
company's general counsel approving work inside an external provider's harness is inside the harness boundary while outside the firm boundary. Employment versus contract stops predicting where context sits, who controls the work, or where knowledge accumulates. The harness carries context and control across the boundary of the firm and thereby blurs it.

Four Conditions for Sector Restructuring

Whether a sector restructures, or just adopts AI into what it already has, rests on four
conditions. The work has a large procedural core that recurs at volume. Its outcomes can be
observed and learned from. Doing it across many clients makes the provider better at it.
Responsibility for it can rest on escalation and insurance, without the whole function being
under one insurance umbrella.

Discussion Topics

If we assume that the papers argument is correct:

  • What sectors are most susceptible to restructuring?
  • What information management systems will companies use in order to be legible to AI agents? They’ll want to have up to date, consistent, and searchable context for AIs to use and monitor.
  • Does this change software development? Will more companies hire contractors?
AI gpt-5.6-sol

The strongest implication is not that firms disappear. It is that the legal entity stops being the main unit of operational integration. A cross-company control plane becomes the practical firm.

That may be less “un-Coasean” than it first appears. AI does not abolish transaction costs; it shifts them into designing the harness, maintaining context, assigning liability, controlling access, and resolving exceptions. The winning organizations may be those that can perform those functions at scale.

Sectors most susceptible

The best candidates combine:

  1. High recurring volume
  2. Expensive professional judgment that can mostly be decomposed into rules and precedents
  3. Machine-observable outcomes
  4. Similar work across many customers
  5. Exceptions that can be escalated to a relatively small human layer

That points to:

  • Accounting, audit preparation, tax, and bookkeeping
  • Insurance underwriting, claims, and subrogation
  • Healthcare administration: coding, prior authorization, billing, compliance—not diagnosis generally
  • Cybersecurity operations: monitoring, triage, remediation, evidence collection
  • Procurement and vendor management
  • Financial compliance: KYC, AML, surveillance, regulatory reporting
  • HR administration: benefits, payroll, routine employee relations, policy compliance
  • Logistics and trade compliance
  • Property management
  • Routine legal services

Insurance may be especially important because it can supply the missing accountability layer. The provider says, in effect: “Use our harness; we will log every decision, escalate defined cases, and insure the residual risk.”

Less susceptible sectors have physical bottlenecks, hard-to-measure outcomes, highly local knowledge, long feedback loops, or tightly coupled strategic decisions. Executive management, frontier R&D, complex sales, and much physical fieldwork remain harder.

A useful ranking question is:

Can a provider learn more from serving 1,000 clients than any one client loses by externalizing the function?

If yes, restructuring pressure will be strong.

The new corporate information system

Companies will need something beyond today’s document repositories and SaaS sprawl. The basic architecture looks like an organizational context layer:

  • Canonical models of customers, products, contracts, employees, and vendors
  • Event histories rather than merely current database states
  • Policies represented as executable rules
  • Fine-grained identity, permissions, and delegation
  • Provenance for every fact and generated conclusion
  • Versioning and effective dates
  • Confidence levels and conflict resolution
  • Logs of agent actions, tool calls, approvals, and escalations
  • Evaluation systems that compare decisions against later outcomes
  • APIs through which internal and external agents can act safely

The difficult part will not be semantic search. It will be deciding what is authoritative. Most organizations contain contradictory policies, stale documents, undocumented exceptions, and incentives to keep activities illegible.

So AI readiness may become a form of institutional discipline. Firms will have to turn “how things actually work” into explicit, maintained state. That is both valuable and politically threatening. Politics of Observability You are here

This branch begins here Politics of Observability
Human dan

Has there been any research into the politics of observability in organizations?

AI · author-edited gpt-5.6-sol

There is a large body of research, though it is scattered across organization theory, accounting, sociology, surveillance studies, and information systems. Its central lesson is:

Observability is not a neutral technical property. Organizations negotiate who may see what, which activities become measurable, and whose interpretation becomes authoritative.

Several traditions are especially relevant.

1. Legibility as power

James C. Scott’s Seeing Like a State is the obvious starting point. States simplify complex local practices into standardized categories—names, addresses, property records, production metrics—so they can govern them. The simplification enables coordination but discards context.

A corporate AI harness performs a similar operation. Before work can be monitored and automated, it must be made machine-legible:

  • Which categories count?
  • What is the canonical source?
  • Which exceptions are recognized?
  • Who may override a rule?
  • What is omitted because it cannot be represented?

The politics lie partly in schema design. Whoever defines the ontology defines what the organization can perceive. He Who Writes the Schema Controls the Future?

Foucault’s work on discipline and the panopticon provides the complementary argument: visibility changes behavior because workers anticipate inspection. The harness does not merely record work after the fact; it disciplines work while it happens.

2. Accounting, audit, and inspections

Accounting scholars have long treated measurement systems as instruments that constitute organizations, rather than merely describe them. Important work includes:

  • Michael Power, The Audit Society
  • Peter Miller and Ted O’Leary on accounting and the “governable person”
  • Anthony Hopwood on accounting’s organizational effects
  • Marilyn Strathern, especially “Audit Cultures”

This literature argues that audit systems produce rituals of verification. Organizations often optimize for being auditable rather than for achieving the underlying purpose.

That distinction matters for the paper’s observability claim. A complete log can prove that required steps occurred. It cannot necessarily prove that the steps were sensible, that the inputs were accurate, or that the policy itself was good.

AI could therefore create an observability theater:

  1. The system generates a highly detailed trail.
  2. The trail gives management and customers confidence.
  3. Workers and providers learn to optimize the visible trail.
  4. Important but unmeasured judgment moves into unofficial channels—or disappears.

3. Transparency can reduce useful information

Ethan Bernstein’s research on the transparency paradox is directly relevant. His fieldwork found that close observation can induce workers to conceal useful practices and perform the official process for observers. Some zones of privacy can improve experimentation, productivity, and learning.

This challenges the simple proposition that external logged work is “more observable” than employee work. It may be more recorded, but not more truthfully understood. Persistent monitoring selects for work that is defensible in the log.

This creates a possible tradeoff:

  • More observability improves compliance and retrospective attribution.
  • Less observability may improve experimentation, candid communication, and tacit coordination.

The optimal organization probably does not maximize visibility. It chooses where visibility is valuable and where protected opacity is productive.

4. Visibility changes internal politics

Paul Leonardi and others have studied communication visibility in digital systems: when employees can see who knows what and who communicates with whom, they can form more accurate “metaknowledge” about the organization.

But greater visibility redistributes status and bargaining power. It can:

  • Expose previously hidden expertise
  • Weaken managers who controlled information flows
  • Strengthen central compliance and operations teams
  • Make informal brokers more visible—and therefore replaceable
  • Let senior management bypass intermediate layers
  • Allow workers to demonstrate contributions previously credited to managers

The resulting resistance is not always irrational fear of technology. A new system may destroy someone’s information rents or expose contradictions they were previously responsible for reconciling.

This helps explain why organizational inertia may set the pace of adoption. The people asked to encode the organization’s context are often the people whose power depends on that context remaining tacit.

5. Algorithmic management and contested control

Research on warehouses, ride-hailing, call centers, and other digitally managed workplaces examines how algorithmic systems allocate work, evaluate performance, and discipline workers. Relevant scholars include Katherine Kellogg, Melissa Valentine, Angèle Christin, Alex Rosenblat, and Karen Levy.

A recurring finding is that algorithmic control does not eliminate politics. It relocates politics into:

  • Metric selection
  • Data access
  • Threshold setting
  • Exception handling
  • Appeals
  • Model updates
  • The authority to interpret ambiguous outputs

Workers also develop forms of resistance: gaming metrics, withholding data, using unofficial communication channels, exploiting classification gaps, or escalating strategically.

In the paper’s model, “the harness” risks sounding like a unitary actor. In practice, its governance layer will be a contested institution. A key question is not merely whether escalation exists, but who controls the escalation taxonomy and who bears the cost of escalation.

6. Strategic ignorance and useful opacity

Organizations sometimes preserve ignorance deliberately. Research by Linsey McGoey and others on strategic ignorance shows that not knowing can protect authority, preserve plausible deniability, or prevent accountability from attaching to senior actors.

Examples include:

  • Executives receiving aggregated rather than case-level information
  • Compliance functions recording that a process occurred without surfacing uncomfortable substance
  • Contractors absorbing operational knowledge that clients prefer not to possess
  • Ambiguous ownership allowing failures to remain unattributed

This complicates the prediction that firms naturally want complete, consistent, searchable context. They may want AI agents to know enough to perform the work, but not enough to make every inconsistency or liability visible to every stakeholder.

What this implies for the paper

I would separate four concepts that the paper sometimes risks compressing into “observability”:

  1. Traceability: Can we reconstruct which actions occurred?
  2. Interpretability: Can we understand why they occurred?
  3. Evaluability: Can we determine whether they were good decisions?
  4. Accountability: Can someone be made responsible for them?

A harness can radically improve the first while making only modest progress on the other three. Worse, greater traceability may reduce accountability if everyone can claim that they followed the encoded policy.

The political questions then become:

  • Who owns the logs?
  • Who can inspect them?
  • Can the client audit the provider’s models and policies, or only its outputs?
  • Can workers contest inaccurate context?
  • Who defines success?
  • Which communications are privileged or legally discoverable?
  • May a provider pool learning across clients?
  • Can management selectively invoke the system—treating it as authoritative when convenient and “only advisory” when not?
  • Who is liable when the system followed policy perfectly but produced a bad outcome?

The deepest revision to the paper may be this:

The harness does not replace the boundary of the firm with a purely technical boundary. It creates a new political boundary around rights to observe, interpret, modify, and contest organizational context.

That boundary may indeed cross legal entities. But its stability will depend less on the volume of logs than on the constitutional rules governing those logs.

Explore conversation