We tend to preserve the arrival.

The answer. The opinion. The final output.

But the work happened before that.

The endpoint is a compression.

The work happens while the answer is still uncertain.

Organizations already have ways to locate the work.

A chart is what we already know: procedure, standard, precedent.

A lookout is attention: someone watching this case, now.

A sounding is a measurement of the water we are actually in.

We do not need everything.

We need what is most important now, in this particular work.

Assurance is not certainty about everything ahead.

It is enough orientation to make the next move defensible.

Agents change how the work is done, and how it moves through the organization.

The instrument does not decide where the voyage should go.

But it can change what the crew is able to see.

What the instrument makes visible—and what it does not—is part of its design.

That view has to be tested.

The people answerable for the work still need to know

what is established, what remains unresolved, what changed, and where attention is needed.

Some readings locate the work.

Some reveal that the chart itself is incomplete.

All should preserve the evidence the next decision depends on.

A harbor is a bounded place to test the whole arrangement:

crew, instruments, and procedures, working together.

But the goal is not to become perfect at staying in the harbor.

The real work begins when conditions stop repeating.

  • Is the course wrong?
  • Is the instrument wrong?
  • Is the chart wrong?
  • Have conditions changed?
  • Are we beyond the map?

These are not settled before the work.

They are answered while it is under way.

Navigational Assurance

Before We Can Value the New Work

In consequential work, the final output rarely ends anything.

An audit opinion becomes a position from which investors, boards, and regulators act. A risk classification becomes a position from which an organization assigns controls. A validated finding becomes a position on which the next decision depends.

What looks like an arrival for one part of the organization becomes an instrument for another: the work establishes a position, the opinion expresses it, assurance is the judgment that others may navigate from it.

Agents complicate that relationship, but they may also enlarge it. As instruments within the work, they can make visible patterns, relationships, and changes that no individual could hold in view. A new reading can allow us to see farther.

But seeing farther is not the same as knowing how to navigate with what we see.

Before a reading can support a consequential position, the organization must be able to establish what it is a reading of, which evidence and expectations produced it, under what conditions it holds, and what it may have missed.

That is what the harbor is for.

A benchmark can characterize an instrument against a fixed reference. A harbor is a validation environment for the whole arrangement of people, instruments, procedures, and evidence.

To be useful, it needs a declared purpose: the position or decision the arrangement is expected to support. It needs scenarios that do more than repeat expected cases: does a late document that changes the conclusion reach the person expected to sign? It needs visibility into what evidence the instruments received, what they changed and what reached professional attention. And it needs a record of what the validation established, what remained unresolved, and where its conclusions stop.

The harbor is not only a gate placed in front of a destination we have already chosen. It is a place to follow unexpected readings and compare arrangements: to learn what new instruments allow us to perceive, how their readings relate to changing conditions, and whether the positions produced by the work can be relied upon. Validation produces evidence for those judgments; it does not make them or sign the opinion.

Some of what we learn will make an existing course more reliable. Some may reveal destinations that were not previously visible. Either way, the purpose is not to become excellent at staying in the harbor. It is to learn enough about the new work arrangement and instruments to take them into open water.

And to recognize when their readings no longer hold or the chart must be revised.

That is where the question of value begins: not with the harbor itself, but with what the organization becomes able to do beyond it.

Outputs measure activity. Costs measure expenditure. Speed measures movement. None tells us whether the organization can navigate from what the work has produced, or whether a new instrument has genuinely allowed it to see farther.

Before we can value the new work, we need to know whether it leaves us better able to navigate.

Our guides

  • Herbert A. Simon

    Bounded rationality, search, satisficing.

    A chooser with finite attention does not optimise over the whole space. It bounds the problem until a decision can be made at all.

    “A Behavioral Model of Rational Choice”, Quarterly Journal of Economics, 1955. Also Models of Bounded Rationality, MIT Press, 1982.

  • Edwin Hutchins

    Distributed cognition.

    Navigation is not performed in one head. It is carried by the crew, the instruments, the chart, and the procedures together.

    Cognition in the Wild, MIT Press, 1995.

  • Jens Rasmussen

    Adaptive work and boundaries.

    Work does not stay where the procedure drew it. Under pressure it migrates toward the boundary of safe operation, and the boundary is rarely marked.

    “Risk management in a dynamic society: a modelling problem”, Safety Science, 1997.

  • David D. Woods

    Resilience, surprise, adaptive capacity.

    Every model of a system is bounded, and surprise arrives from outside the bound. What matters then is the capacity to stretch.

    “Four concepts for resilience and the implications for the future of resilience engineering”, Reliability Engineering & System Safety, 2015.

  • Terence Tao

    Inquiry, and what the path produces beyond the answer.

    What a piece of work produces is not only the result. The path leaves behind technique, intuition, and questions the statement of the answer does not carry.

    “What is good mathematics?”, Bulletin of the American Mathematical Society, 2007.

Earlier essays from Reins AI →