What Has AI Revealed Was a Waste of Time?

The machine did not devalue work. It exposed work that was never where the value lived.

Framing the Question

AI busywork is becoming easier to see because the machine is fast at the very things many workplaces quietly rewarded: producing an acceptable first draft, restating available information, filling templates, and making routine communication look finished. The question is uncomfortable because time spent is often mistaken for value created. When a tool compresses an hour into a minute, it does not prove that the person was useless. It asks whether the hour had been spent on the right part of the job.

The Direct Answer: Work That Only Imitated Value

AI has made one category of work especially hard to defend: predictable production performed as though production itself were expertise.

That includes writing the fifth variation of a standard customer reply, converting meeting notes into a familiar summary, manually reformatting information already present in systems, producing routine code scaffolding, and polishing a document before anyone has decided whether the argument is right.

The waste was not typing, coding, or writing. Those can be valuable crafts. The waste was requiring a skilled person to repeatedly manufacture a foreseeable output when the real value lay elsewhere: understanding the exception, setting the standard, making the choice, checking the risk, or taking responsibility for the consequence.

This is the distinction AI has made visible:

Output labor creates a draft.
Judgment labor decides whether that draft should exist, can be trusted, and changes anything.

The Clue Hidden in Who Benefits Most

Consider customer support. In a large workplace study later published in The Quarterly Journal of Economics, access to a generative-AI assistant increased issues resolved per hour by 15 percent. Earlier working-paper findings showed especially large gains for newer or lower-skilled agents, suggesting that the tool helped transfer patterns of strong performance to people still learning the role. s more revealing than “AI makes people faster.” It suggests that a portion of the old process consisted of making each new employee independently rediscover good phrasing, troubleshooting paths, and conversational moves that the organization already knew.

Some learning is indispensable. Requiring thousands of people to repeatedly reconstruct the same usable response is not development; it is a knowledge-transfer failure.

AI did not prove that support workers were wasting time by helping customers. It exposed the waste in forcing them to hunt for standard moves while an upset customer waited.

When a Blank Page Was Mostly a Toll Booth

Software development offers a different example. In a controlled experiment, developers with GitHub Copilot completed a specified JavaScript HTTP-server task 55.8 percent faster than developers without it. The task was narrow; it does not show that software engineering is solved. It does suggest that recalling and assembling standard implementation patterns can be an expensive toll booth on the way to the work that requires deeper understanding. lem appears outside code.

A finance analyst spends forty minutes making last month’s slide deck match this month’s numbers. A recruiter rewrites the same interview-scheduling email for the twentieth time. A manager spends Sunday evening turning six project updates into one bland status memo that few people will read closely.

None of these people is lazy. The system has asked them to demonstrate diligence through friction.

AI raises an awkward question: if the recurring artifact can be generated in seconds, why did the organization measure commitment by how long it took to make?

The Friction Audit

A useful QuestionClass test is the Friction Audit. Take a repeated task and ask three questions:

Predictability: Would two competent people produce largely the same acceptable output?

Consequence: If the first draft is wrong, is the harm low and easily caught, or high and difficult to reverse?

Residual judgment: After AI produces a draft, what decision still requires a human who understands the stakes?

When predictability is high, consequences are containable, and residual judgment is clear, spending human hours on first-pass production is probably avoidable waste. Let AI draft; require people to define, inspect, decide, and own.

But when the consequence is high or the task depends on facts the tool cannot safely establish—diagnosing an illness, interpreting a contract, making a firing decision, promising a client an outcome—the fast draft can conceal more work than it removes.

The question is not, “Can AI do this?” The question is, “What kind of human attention does this task still deserve?”

The Seductive Mistake: Confusing Faster With Unnecessary

This is where the question needs resistance. AI has not made all slow work look foolish. Sometimes doing the work is how people build the very judgment they later need to supervise AI.

A study with Boston Consulting Group professionals described AI’s jagged technological frontier: assistance improved performance on some knowledge-work tasks yet worsened performance on others that looked similarly approachable. The dangerous lesson is not “never use AI.” It is that polished speed can make users less alert precisely where judgment is most needed. rney who never struggles through a contract may become efficient at accepting plausible clauses without sensing the trap. A new programmer who only accepts completions may ship quickly without understanding failure modes. A young manager who delegates every uncomfortable message to a model may never learn how trust is repaired in a difficult conversation.

Removing drudgery is useful. Removing apprenticeship is expensive.

The better design is not “automate everything tedious.” It is: automate repetition after identifying what practice people still need in order to judge the output.

A Sharper Question

Instead of asking:
“What has AI made perfectly clear to have been a waste of time?”

Ask:
“Which repeated tasks consume skilled attention without building judgment, protecting trust, or changing the decision?”

The original question can tempt us into contempt for yesterday’s work. The sharper question separates worthless friction from necessary learning and accountable judgment.

What to Do With This

In your next team meeting, put one recurring deliverable on the screen: the weekly status memo, proposal summary, support-response draft, research recap, or code documentation update.

Ask the Friction Audit questions aloud.

If the artifact is predictable and low-risk, stop discussing whether AI is impressive. Run a controlled change:

1. AI prepares the first version.
2. A named person verifies facts, exceptions, and tone.
3. The team records time saved and errors caught for four cycles.
4. The work is redesigned only after seeing what actually improved.

Then make one deliberate reinvestment. Use recovered time to call the frustrated customer, test the assumption behind the dashboard, coach the new hire through an unusual case, or ask whether the project should continue at all.

Without reinvestment, automation merely produces more output. With it, automation buys attention for better decisions.

Bringing It Together

AI has made it clear that much of what looked like professional effort was really a tax on attention: generating predictable artifacts, locating already-known patterns, and translating information between formats. Yet it has also clarified what was never waste: learning the terrain, questioning the brief, spotting the exception, earning trust, and taking responsibility.

The most useful response to AI is not to celebrate speed or mourn effort. It is to ask, with much greater precision, what deserves a human mind. Build that habit one question at a time with QuestionClass’s Question-a-Day at questionclass.com.

Bookmarked for You

These books help distinguish productive tools from systems that quietly train people to confuse activity with value.

The Second Machine Age by Erik Brynjolfsson and Andrew McAfee - A useful foundation for understanding how digital tools change not only productivity, but the division of work between people and machines.

Working Backwards by Colin Bryar and Bill Carr - Amazon’s operating practices show why clear decisions and customer outcomes matter more than the internal theater of producing documents.

The Checklist Manifesto by Atul Gawande - A sharp exploration of how routine process can support expert judgment without pretending that procedure replaces expertise.

QuestionStrings to Practice

A QuestionString turns a vague reaction—“this task feels pointless”—into a sequence that reveals whether the problem is repetition, risk, or misplaced human attention.

The Attention Recovery String
For when a recurring task suddenly looks automatable:

“Which part of this task repeats with little variation?” →
“What could go wrong if AI produced the first draft?” →
“What judgment still requires context, experience, or accountability?” →
“Which skill would weaken if we removed the task completely?” →
“Where should the recovered attention go instead?”

Use this string when reviewing recurring reports, communication workflows, research summaries, customer responses, or internal documentation. It prevents two bad decisions: protecting pointless friction and automating away the practice people still need.