I do not remember who first showed me how to heave-to under sail, but I remember how strange it felt. Everything I had learned until then had been about making the boat go. Trim the sails. Watch the wind. Keep your speed up. Hold your course. Then, in the middle of San Francisco Bay, someone calmly suggested that we stop.

Not anchor or turn around, but simply stop trying to make progress for a while. By balancing the sails against the rudder, a sailboat can settle into a slow, controlled drift. The boat remains stable and manageable while the crew studies the chart, waits for weather, repairs something, eats, rests, or simply thinks. No one has abandoned the voyage. They have created enough stability to decide what the next part of it should look like.

Organizations rarely give themselves the same opportunity. When a problem appears, the instinct is usually to add something: another system, another report, another committee, another approval, another consultant, or another layer of technology. More recently, the answer is often artificial intelligence. Any of these may be useful, but they are frequently introduced before the organization has paused long enough to understand the problem.

A New Kind of Interface

For most of the history of computing, people had to learn the language of the machine. We learned commands, menus, forms, queries, reports, dashboards, and the particular logic of each application. Even simple questions often required knowing which system contained the information, which report exposed it, and which person understood what the result actually meant.

Natural language changes that relationship. We can now ask a computer a question in ordinary words and receive a useful response in seconds. Instead of navigating through applications, folders, reports, and search results, we can begin with the question itself. That is a profound change in the way people interact with information.

It is tempting to imagine that easier access to answers will resolve the information problems that organizations have struggled with for decades. It will not. Natural language changes the interface, but it does not automatically improve the knowledge behind it. A confident answer can still be incomplete, outdated, poorly sourced, or simply wrong.

An Old Question Remains

During my career, I have watched one generation of technology after another promise to solve the problem of organizational knowledge. Mainframes gave way to client-server systems, then enterprise software, data warehouses, business intelligence, cloud computing, data lakes, digital twins, and now artificial intelligence. Each generation advanced the state of the art and made things possible that had previously been difficult or impractical.

None eliminated the question that has followed me into almost every organization I have worked with:

Can we trust what we know?

It sounds like a technical question, but it almost never is. Over the years I have worked with engineers, operators, scientists, accountants, planners, regulators, executives, software developers, GIS specialists, laboratory managers, maintenance crews, and field inspectors. They used different systems, spoke different professional languages, and cared about different outcomes. After a while, however, the conversations began to sound remarkably familiar.

Someone wanted to know why costs were increasing. Someone else wanted to know whether a capital project had actually improved performance. Another person questioned why two reports produced different numbers. A manager wondered why staff still maintained spreadsheets after millions of dollars had been invested in enterprise systems. Eventually, someone would ask why it was so difficult to get a straight answer. That was usually the point when the real work could begin.

Answers Depend on What Lies Beneath

An artificial intelligence system can summarize documents, but it cannot decide which document should have been trusted in the first place. It can identify a pattern in operational data, but it may not know that a sensor was out of calibration. It can explain a regulation, but it cannot determine whether an organization has followed it unless the relevant evidence exists, can be found, and can be traced to a reliable source.

The quality of the answer depends on the knowledge environment beneath it. That environment includes systems and documents, but it also includes definitions, ownership, history, assumptions, quality controls, operating practices, and the experience of the people who understand how the organization really works.

Artificial intelligence does not remove the need for that foundation. It exposes it. When the underlying knowledge is coherent and well maintained, natural language can make it remarkably accessible. When the foundation is fragmented or poorly understood, the same technology can produce faster confusion.

Complexity Is Usually Additive

Organizational solutions tend to be additive rather than subtractive. A broken process receives a workaround. The workaround becomes a spreadsheet. The spreadsheet produces a report. The report creates a meeting. The meeting produces an action log. Eventually, another system is purchased to manage the complexity created by the earlier systems.

Anyone who has owned an older sailboat will recognize the pattern. Previous owners add switches, pumps, instruments, wires, breakers, and labels. Each addition probably made sense at the time. Years later, the boat carries the accumulated decisions of people who are no longer there, and no one is entirely sure what can safely be removed.

Organizations age in much the same way. They accumulate applications, interfaces, policies, exceptions, approvals, reports, and committees. Complexity becomes easier to add than to understand, while subtraction begins to feel risky. Artificial intelligence may help people navigate that complexity, but it can also become one more layer placed on top of everything that came before.

A faster answer does not help if it comes from a system no one understands.

The Crew Is Part of the System

Technology discussions often treat organizations as though people were interchangeable components in a machine. If performance is poor, the assumed remedy is a better instrument, a faster platform, or a redesigned process. Less attention is given to the condition of the people expected to use them.

A crew may not be sailing well because the boat lacks equipment. They may be tired, hungry, unclear about the plan, or assigned to the wrong tasks. The best helmsman may be below completing paperwork. A skilled engineer may be buried in routine administration. A manager may be performing quality control because responsibility was never established elsewhere.

Organizations experience the same drift. Tasks move toward whoever is willing, familiar, or available rather than toward the person with the right skill, authority, or perspective. A new system may reduce some of the burden, but it may also give an already tired crew one more instrument to monitor, another workflow to maintain, and another source of information to reconcile.

Navigation Before Acceleration

A navigator is surrounded by information. Wind direction, tide, current, depth, weather, position, traffic, equipment, provisions, destination, and the condition of the crew all matter. No single observation determines the next move. Navigation is the practice of combining incomplete and sometimes conflicting information into a sound decision.

Organizations face the same challenge. Before they automate, accelerate, or add another layer, they need to understand where they are, what they are carrying, what can be removed, and whether the crew is ready for the next leg. Sometimes the right response is a new instrument. Sometimes it is a simpler process. Sometimes the crew needs training, clearer roles, food, or rest. Sometimes the work has simply been assigned to the wrong skill set.

Artificial intelligence gives us a powerful new way to ask questions and explore what an organization knows. It does not relieve us of the responsibility to judge the answer, understand its source, care for the crew, and decide whether the existing course still makes sense. The technology will continue to change. The discipline of navigation will not.

Next Bearing

Before intelligent systems can simplify an organization, we need to understand why complexity is so easy to add and so difficult to remove. The next Bearing considers how vessels and organizations gradually become heavier, even when every individual addition once seemed reasonable.