AI Made It Easier. That Does Not Make It Worth Less.
2026-05-18 · 6 min read · Janaina Maia
One of the strangest arguments around AI-assisted work is that if AI made something easier, the result must be worth less.
I understand the instinct. We are used to reading effort as a signal of value. If something took ten hours, it feels more serious than something that took twenty minutes. If someone struggled, revised, stayed late, and suffered a little, we assume more craft went into the outcome.
But effort is a very poor proxy for value. It always has been.
The real question is not “how hard was this to make?” The better question is: did it solve the right problem, with good judgement, in a way someone can trust?
This argument is older than AI.
Every time a technology reduces the visible labour of making something, people worry that the work has become less legitimate.
The printing press made books easier to reproduce. Before movable type, manuscripts had to be copied by hand. Gutenberg’s system made it possible, as Britannica summarises, to manufacture large numbers of books at relatively low cost, making printed material available to a wider audience. The value did not disappear because copying became easier. The value moved: from hand-copying every page to editing, publishing, distribution, literacy, interpretation, and institutional trust.
Photography created a similar panic. If a camera could capture an image faster than a painter, was the image still art? The answer, historically, was not to declare photography worthless. It was to develop a new language of composition, timing, framing, documentary evidence, and artistic intent. The Metropolitan Museum of Art describes William Henry Fox Talbot’s early photography as a “wholly new way of making pictures” that served botanists, historians, travellers, and artists. The mechanical process did not remove human judgement. It changed where judgement lived.
Walter Benjamin made this tension famous in “The Work of Art in the Age of Mechanical Reproduction”. Reproduction changed the aura of art, but it also changed access, politics, distribution, and the relationship between the public and the work. The old value system was disrupted. A new one formed.
AI is another version of the same pattern. When production becomes easier, people mistake the disappearance of one kind of labour for the disappearance of value itself.
Ease changes the value chain. It does not erase it.
Generative AI is genuinely reducing effort in some parts of knowledge work. That is not a moral failure. It is the point of tools.
In a preregistered experiment with 444 college-educated professionals, Shakked Noy and Whitney Zhang found that access to ChatGPT reduced time spent on professional writing tasks by 37% and improved evaluator grades by 0.45 standard deviations. Their paper, “Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence”, also found that the work shifted away from rough drafting and toward idea generation and editing.
That last point matters. AI did not simply make people “do less.” It changed the shape of the task.
Another study, “Generative AI at Work” by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond, found that an AI conversational assistant increased customer support productivity by 14% on average, with a 34% improvement for novice and lower-skilled workers. The interesting finding is not only speed. The authors suggest the tool helped disseminate the practices of more experienced workers.
In other words, AI can compress access to patterns that used to take years to absorb. That does not make expertise irrelevant. It makes expertise more important as the thing that selects, adapts, reviews, and owns the output.
The valuable work is moving upstream and downstream.
When AI can generate a first draft, the value of typing the first draft goes down. That is uncomfortable if your identity is attached to typing the first draft. But it is not the end of professional value.
The valuable work moves upstream:
- Framing the right problem.
- Knowing what context matters.
- Choosing the right constraints.
- Setting the quality bar.
- Knowing what good looks like before the model produces anything.
And it moves downstream:
- Evaluating whether the output is true.
- Editing for audience, tone, ethics, and consequence.
- Checking assumptions and sources.
- Deciding what should not be shipped.
- Taking accountability for the final result.
That is why “AI made it” is not enough information. AI made what? Under whose direction? With what source material? Reviewed by whom? For what audience? With what consequences?
A weak person with AI can produce a lot of weak work faster. A strong person with AI can explore more options, test more angles, and spend more time on judgement. The tool amplifies the operating model around it.
We need to stop romanticising inefficient labour.
There is a difference between craft and friction.
Craft is the judgement that improves the work. Friction is the effort that merely consumes time. We confuse them constantly.
Writing ten versions of a paragraph can be craft if each version sharpens the point. Reformatting a deck for three hours is usually friction. Searching through a messy folder for the right file is friction. Manually turning meeting notes into a first summary is often friction. Waiting for someone to create the first draft before the real conversation can begin is friction.
If AI removes friction, good. That gives us more room for the work humans should actually be doing: deciding, questioning, challenging, refining, and taking responsibility.
The danger is not that AI makes work too easy. The danger is that teams will use AI to create more mediocre outputs instead of better decisions.
The design implication: show the judgement, not just the output.
If people distrust AI-assisted work because it looks easy, part of the answer is better product design.
AI tools should make the human contribution more visible. Not performatively, as if we need to prove we suffered. But practically, so others can trust the result.
For AI-assisted work, I want to see:
- The brief or intent behind the output.
- The sources or inputs used.
- The alternatives considered.
- The human edits and approvals.
- The risks, caveats, and open questions.
- The person accountable for shipping it.
This is especially important in enterprise products. If an AI generates a report, recommendation, workflow, or customer-facing message, the interface should not only show the final artefact. It should show enough of the decision trail that a human can own it.
That is where trust will come from. Not from pretending AI was not involved, and not from treating AI output as automatically valuable. Trust comes from visible judgement.
My rule of thumb.
If AI made something easier, do not ask whether it still has value. Ask where the value moved.
Sometimes the answer will be disappointing. The work really was mostly manual production, and AI exposed that. But often the answer is more interesting: the value moved to taste, strategy, context, review, ethics, and accountability.
That is not less human. It may be more human.
The future will not reward people for doing everything the hard way. It will reward people who know which parts should be easy, which parts must remain careful, and which decisions still require a human name attached to them.
AI does not make work worthless because it makes work easier. It makes weak judgement more visible. It also gives strong judgement more leverage.