Productive laziness and the new operating system for work

AI can remove repetitive work at extraordinary speed. It can also generate more work than anyone needs. The operator’s task is to know the difference.

There is a particular kind of work that has always irritated me: a capable person copying the same information into the same format because no one has stopped to redesign the process.

I used to think this was impatience. Over time, I came to see it as a useful operating instinct.

Productive laziness is a low tolerance for work that should no longer exist.

It is not an unwillingness to work. Quite the opposite. Removing work responsibly often requires understanding it more deeply than simply repeating it.

You need to know why the work exists, what outcome it supports, where judgment enters and what could go wrong if the process changes. Only then can you decide whether to simplify it, automate it or remove it entirely.

AI has made that distinction harder and more important.

It can now remove effort at extraordinary speed. It can also accelerate the wrong process, preserve a weak assumption and produce more work than anyone has time to judge.

The question is no longer simply how much work we can automate.

It is which work should disappear, which should become faster and which must remain human.

The cost of execution fell

The first generation of operational improvement was largely about digitisation.

Paper became data. Conversations became workflows. Spreadsheets became systems. Information became easier to store, retrieve and move.

The next step was automation. If a process was sufficiently stable and repeatable, software could perform parts of it without continuous human intervention.

AI changes the economics again.

It can draft an analysis, structure a meeting, compare documents, prepare a decision, explore a market, build a prototype or turn an unstructured thought into something another person can review.

Work that once required several uninterrupted days can sometimes be compressed into hours.

My first reaction to this was to focus on speed.

The faster I could produce, the more I could accomplish. A shorter distance between an idea and a working version meant that more ideas could be tested. Problems did not need to wait for an available specialist before they could be explored.

But the more production accelerated, the clearer another constraint became.

Execution was no longer the main bottleneck. My own attention was.

Every additional output still needed to be understood. Every option needed to be assessed. Every useful draft created another decision about what should happen next.

Cheap production can create expensive review.

At a certain point, generating more becomes less useful than deciding better.

Faster can still be wrong

AI can build a dashboard before anyone has defined the decision it should support.

It can write a procedure before the underlying process has been challenged.

It can produce ten strategic options when the team needs the discipline to commit to one.

It can automate a handoff that should not exist in the first place.

This is why productive laziness must begin before automation.

The first question is not:

Can this be automated?

It is:

Should this work exist at all?

Automation does not correct a weak operating model. It allows the weakness to travel faster.

I have experienced this directly while building software with AI.

I could create functions, modules and working features far faster than my technical capability would previously have allowed. The output could look convincing and work in the immediate case.

But speed did not give me the accumulated judgment of a senior developer.

I could still make weak architectural choices. I could solve a local problem in a way that created complexity elsewhere. I could build something functional without recognising the cost of maintaining it.

AI expanded what I could produce. It did not automatically expand my ability to recognise every flaw in what I had produced.

The same applies to leadership and operations.

A model can help prepare a financing analysis, but it does not carry responsibility for the assumptions.

It can identify patterns in customer feedback, but it does not decide which customers the company should build for.

It can prepare a board paper, but it does not own the consequences of the decision.

It can accelerate the work around judgment. It cannot make accountability disappear.

This makes expertise more valuable, not less.

When production is slow, expertise is partly expressed through the ability to create the output. When production becomes fast, expertise is increasingly expressed through the ability to frame the problem, test the result and recognise what should never be accepted.

Not all work deserves the same response

I now see four possible responses to work.

Remove what serves no useful purpose.

Some work exists because of habit, an outdated control or a process designed for circumstances that no longer apply. Automating it would only make an unnecessary activity more efficient.

Automate what is repeatable and bounded.

Where the inputs, rules and acceptable outcomes are sufficiently clear, repeated human effort adds little value. These are good candidates for automation, provided that exceptions and escalation paths remain visible.

Augment work that benefits from preparation but still requires judgment.

Research, synthesis, meeting preparation, analysis and scenario development can all become faster. AI can prepare the ground, but a person still needs to interpret the evidence and decide what it means.

Keep consequential decisions human-owned.

The greater the effect on people, capital, safety, reputation or strategic direction, the clearer the human ownership must be. AI may inform the decision, but it should not blur who is accountable for making it.

This is the practical meaning of productive laziness.

It is not maximum automation. It is disciplined allocation of human attention.

The purpose is not to eliminate effort indiscriminately. It is to stop spending skilled attention on work that does not deserve it, so that more attention remains for the work that does.

What remains the operator’s work

As execution becomes cheaper, the role of the operator changes.

The leverage moves from producing every step to designing the work, setting its boundaries and remaining responsible for the result.

That means defining:

• the objective the work should serve
• the evidence the system may rely on
• the constraints it must respect
• the decisions it may prepare
• the decisions a person must make
• the standard the output must meet
• the conditions under which the process should stop or escalate

These boundaries matter because AI is naturally good at continuing.

It can keep proposing, expanding, rewriting and exploring. But an organisation does not benefit from endless possibility. It benefits from useful work reaching a decision, an owner and a conclusion.

The operator must therefore provide something the system cannot establish on its own: direction.

Direction determines what deserves attention.

Structure determines how that attention becomes action.

Judgment determines whether the result is good enough.

Accountability ensures that someone still owns what happens next.

Without these elements, AI can create the appearance of progress while increasing organisational noise.

With them, it can remove a considerable amount of friction from serious work.

The work worth keeping

I still believe in productive laziness.

I believe repeated manual work should make us uncomfortable. I believe processes should be questioned before they are repeated. I believe capable people should spend less time transferring information and more time interpreting it.

But AI has made the philosophy more demanding.

It is easy to automate a visible task. It is harder to understand the system around it.

It is easy to produce more. It is harder to decide what deserves to exist.

It is easy to remove effort. It is harder to preserve the judgment, learning and accountability that the effort may have contained.

Productive laziness is therefore not the pursuit of minimum effort.

It is the refusal to spend human attention on work that no longer deserves it.

AI expands what can be removed, automated or prepared. It does not decide what matters, what is safe or who remains responsible.

The operator’s task is not to automate everything.

It is to draw the boundary well.

Some work should disappear.

Some work should become faster.

And some work should remain difficult, because difficulty is where judgment enters.