People are extraordinarily good at holding weak systems together.

In operational work, I keep encountering important activities that appear to function because someone remembers the next step, knows where the current information lives, notices when a checklist has been missed and catches the edge case before it becomes a problem.

The work is functioning. The system is not.

Human flexibility can hide system weakness for a long time. A person fills the gap, learns the unwritten sequence and adapts when reality does not match the process. Because the outcome still arrives, that effort can look like part of the design rather than evidence of what the design is missing.

The hidden support loop

  1. 01The system leaves a gapState, sequence or ownership stays implicit
  2. 02A person adaptsThey remember, interpret and quietly check
  3. 03The work still succeedsThe missing support remains survivable
  4. 04The gap stays hiddenMore work becomes dependent on the person

The successful outcome feeds the same weak design ↺

Break the loop by changing what carries the work

The system carriesstate · sequence · feedback · safeguards
People carryjudgement · care · creativity · exceptions
When human adaptability keeps delivering the outcome, a system can appear stronger than it is. Support makes the routine path explicit without removing human judgement.

Human flexibility can hide system weakness

That flexibility is valuable. People can interpret ambiguity, combine incomplete information and respond to circumstances a process did not anticipate. Those are not defects to automate away.

The problem begins when repeatability, memory, consistency and vigilance are also assigned to people by default. A repeated activity becomes dependent on a particular person knowing what to do, remembering to do it and being present when it matters. Protecting the activity then requires more checking, more documentation and more knowledge held by the same people on whom it already depends.

When that fails, it is tempting to call the result human error. Often the deeper error is that the system required reliable repetition without providing reliable support.

James Reason drew a similar distinction between a person approach to error and a system approach. The first concentrates on forgetfulness or inattention; the second examines the conditions in which people work and the defences that prevent or reduce harm (Reason, 2000). His article concerns clinical safety, where the stakes and safeguards are different from ordinary operational work. The useful connection is not that every mistake is a system failure. It is that blame can stop the investigation one level too early.

System-supported does not mean automated

I used to reach too quickly for the word automation. It implies that the goal is to remove the person from the work. That is not the distinction I am trying to make.

An activity can remain human and still become system-supported. The system can hold the current state, make the next action clear, bring the relevant knowledge into the work, confirm a dangerous choice and expose an exception that needs judgement. Automation may perform some steps, but it is only one form of support.

Lisanne Bainbridge described one of the enduring ironies of automation: automating a process can expand rather than remove the operator's problems, particularly when the person is left responsible for abnormal conditions (Bainbridge, 1983). That paper examined industrial process control, not modern knowledge work. But it gives this distinction a useful edge: removing routine participation can also remove the context needed when judgement becomes most important.

Repetition is therefore a reason to inspect an activity, not a reason to remove a person from it. If the work is repeated, draining, prescribed and does not meaningfully depend on creativity or judgement, I want to know why the routine path still relies on memory and vigilance.

Humans are like water

We tend to follow the easiest available route, particularly in work we perform repeatedly. I do not think that is laziness. It is how people preserve attention in environments that are always asking for more of it.

In six behavioural experiments, Wouter Kool and colleagues found that participants tended to avoid courses of action with greater demands on controlled information processing (Kool and colleagues, 2010). The laboratory tasks do not prove that people always choose the easiest route, and incentives changed the effect. They do support a narrower point: cognitive demand is one of the costs people weigh when choosing how to act.

Habit research adds a different mechanism. Wendy Wood and David Neal describe habits as associations between responses and the contexts in which they have been repeated; once established, features of the context can cue the response without a fresh decision about the goal (Wood and Neal, 2007). If the environment repeatedly cues the workaround, another instruction may be weaker than changing the environment.

If the intended route requires searching through several copies of a document, remembering an invisible rule or working around the tool in front of us, the path of least resistance will eventually separate from the prescribed process. More reminders may temporarily force the two together. They do not change the landscape.

A well-designed process or system changes that landscape. It makes the sustainable, safe route the natural one: the current information is where the work happens, the next step follows from the present state and mistakes become visible before they become irreversible.

Documentation should support the route

Documentation matters. It helps someone understand a process, learn why it exists and recover knowledge that should not live only in another person's head.

But documentation is a reference, not an executing process. It can describe the intended state without showing the actual one. It can list a check without noticing that the check was missed. It can explain a destructive action without preventing it.

Sometimes documentation is the proportionate support, especially while a process is new, rare or still changing. The warning sign is a repeated weakness that accumulates more written instruction while continuing to depend on perfect recall and constant enforcement.

Leave room for people

Not every repeated activity should be reduced to a prescribed path. Care is repeated. Leadership is repeated. Difficult conversations, creative work and decisions between competing values recur without becoming mechanical.

Good system support should make more room for those human qualities, not treat them as inconvenient variation. It should carry routine state and repetition while making it clear when an exceptional circumstance needs a person to depart from the normal route.

This is also why “enables rather than enforces” needs a boundary. Some actions are dangerous enough that a system should prevent them or require deliberate confirmation. Safeguards should be proportionate to the consequence. But if ordinary good practice depends on policing people through every repetition, the system is working against the behaviour it needs.

Enforcement fights human behaviour. Good system design channels it.

Further reading

These books offer related lenses rather than direct proof of the argument. I am including them as a reading list to examine, not as endorsements of every claim they contain.