What Remains After Automation

In the age of AI, the ability to discover and create meaningful differences matters.

When people talk about AI, automation is the word that comes up most often. Writing, translation, summarization, image-making, coding, research, customer support—countless tasks that once had to be handled directly by people are being automated quickly. Companies talk about productivity gains, while individuals look for ways to save time.

Automation clearly matters. But can automation itself become a lasting competitive advantage? If everyone can use similar AI, automate similar tasks, and produce results at roughly similar levels, automation soon becomes a baseline condition. A capability that everyone can use makes life more convenient, but by itself it does not create a distinct difference.

What competitive advantage remains after automation becomes widespread? I think it is the ability to discover meaningful differences within sameness and turn those differences into new value.

1. Automation Raises the Average Level of Capability

The most direct effect of generative AI is that it helps more people produce results above a certain standard, quickly. Expressions, work procedures, and problem-solving methods that once accumulated only in experienced practitioners’ heads are now being distributed widely through AI. With its help, a beginner can write a reasonably natural email, design the structure of a report, generate code, or summarize complex material. Some abilities that once required years of experience are beginning to arrive in the form of a tool.

Empirical research confirms this change. A study by Erik Brynjolfsson, Danielle Li, and Lindsey R. Raymond, analyzing 5,172 customer-support agents, found that the productivity of agents who used generative AI rose by about 15% on average. The effect was especially large for less experienced or lower-skill workers: the number of problems they solved per hour increased by about 30%.[1] The next finding is just as interesting. The researchers also found that lower-skill agents began communicating with customers in ways similar to higher-skill agents after using AI.

AI does not turn everyone into a top expert. It does, however, raise the floor of basic performance quickly. This is a positive change that increases productivity across society, but it also creates a new problem. If anyone can produce polished writing and images, code and proposals, it becomes difficult to hold attention for long on the strength of clean, finished execution alone. As automation spreads, the “well-made average” becomes abundant—and what is abundant loses its scarcity.

Paradoxically, the spread of automation increases the value of abilities that are difficult to automate: a distinctive sense of which problems matter, a distinctive perspective, and judgment.

2. The Risk in the AI Era Is the Homogenization of High Quality, Not Low Quality

The fundamental risk brought by AI is not that the quality of individual outputs will decline. It is that individual outputs may improve while all the outputs together become more alike.

Generative AI learns patterns in expression, structure, images, and ways of thinking from vast bodies of existing data. That lets it produce stable, plausible answers quickly. But as more people use the same models and enter similar requests, their starting points and the structures of their results are also more likely to converge. While smooth sentences, reliable compositions, and familiar images are produced at speed, unfamiliar questions, uncomfortable perspectives, and hard-to-explain individuality may disappear.

The experimental study by Anil R. Doshi and Oliver P. Hauser shows this point. Short stories written with ideas suggested by AI were rated, on average, as more creative, better written, and more entertaining. Yet when the researchers compared the outputs of the group as a whole, the AI-assisted stories were more similar to one another. They interpreted this as a case in which generative AI can raise individual creative performance while reducing collective diversity.[2]

The study reveals a central tension of the AI era. The quality of each work may improve, while the diversity of perspectives produced across the culture may shrink. Every text may be logical, every image polished, and every proposal well organized. But if they begin to resemble one another, we gain quality at the cost of the number of ways we can look at the world.

That is why competitiveness in the AI era does not come simply from making a good result. More important is the ability to decide which of the average paths AI naturally presents to accept, which to question, and where to depart from them.

3. What Does It Mean to Discover a Difference?

People imagine different shapes in the same cloud. One person sees an animal, another a huge city, and another a painting seen long ago or someone they had forgotten. The cloud is the same, but the meaning discovered in it differs for each person.

This difference does not come from eyesight. Because each person has lived through different experiences, memories, interests, and questions, we see the same object within different relationships. Discovering a difference, then, is not finding a correct answer that was already complete and hidden inside a thing. It is closer to making a meaning appear by connecting an object and one’s own experience in a new way.

Seen this way, the central ability people need in the AI era is not only the ability to produce an output. It is the ability to choose a problem, set its context, and judge what the result means. The boundary between what AI can and cannot do is not clear-cut. Of two tasks that look similar, AI may perform extremely well on one and, on the other, be wrong while presenting a plausible line of reasoning. Using AI well therefore means more than entering an instruction and accepting the result. It means deciding when to trust AI, when to doubt it, and where to bring human judgment into the process.

A study of 758 consultants at Boston Consulting Group shows this boundary clearly. Participants using GPT-4 completed tasks within AI’s capability range more than 25% faster, and the quality of their results was more than 40% higher. In complex management tasks designed to fall outside AI’s capability range, however, the group using AI was 19 percentage points less likely to reach the correct answer.[3] The researchers called this boundary, which changes irregularly from task to task, the “jagged technological frontier.”

Expertise in the AI era is not built on tool-use skill alone. We must be able to ask questions such as: Is this problem really worth solving? Whose perspective is missing? What has been smoothed over too neatly? Which exceptions have been erased in the average? When this result is applied to reality, who benefits and who is disadvantaged? Can I take responsibility for this result?

The ability to discover a difference is not a technique for spotting minor discrepancies. It is the judgment to recognize possibilities that do not yet have names within a familiar order.

4. For Deleuze, Difference Is a Principle of Generation, Not the Result of Comparison

Deleuze’s Difference and Repetition was not written to predict the age of AI. It nevertheless offers a useful philosophical lens for interpreting the relationship between automation and creativity.

We usually understand difference by comparing two things. We imagine that A and B exist first, and that we find their differences only after comparing them. In that view, difference is treated as a gap between already existing instances of the same. Deleuze argued that this way of thinking subordinates difference to identity. What matters to him is not a difference found by comparing finished objects, but difference as the generative force that allows things and events to come into different forms.[4]

From Deleuze’s perspective, the world is not a place where a single original is repeatedly copied. It is a place where conditions and relationships change with every repetition, allowing something new to emerge. Seasons repeat each year, but the same spring never returns. When we reread the same book, we find sentences different from those we found before. Even when we ask the same question, the answer changes as time and experience change. Repetition can be a condition in which difference occurs, rather than a copy of the same.

This perspective applies to AI as well. AI is one of humanity’s most powerful repetition devices. It repeatedly generates sentences, transforms images, and tries countless possibilities in a short time. But a high number of repetitions does not automatically create new value. Producing hundreds of pieces of content with the same structure and the same point of view is only an increase in output. Conversely, even if we repeat the same question, changing the conditions, connecting unfamiliar objects, and questioning existing answers can make that repetition generate a new difference.

Whether the repetition AI performs remains simple copying or becomes generative repetition that gives rise to new differences ultimately depends on what humans ask and what they choose.

5. Difference Is Both a Competitive Advantage and the Starting Point of Value

Difference does not always create value. A result that is merely unlike everyone else’s may end up as something strange or uncomfortable. What the market and society require is not difference for its own sake, but meaningful difference.

A meaningful difference has three conditions. First, it reveals a problem that was not sufficiently visible before. Second, it offers a better experience or interpretation than existing approaches. Third, it gives a person’s distinctive perspective a form that other people can understand and use.

Discovering a difference is therefore only the beginning of creation. It becomes social value when we give the discovered difference a concrete form—as language, a product, a service, education, or art. At this point, creativity is not a mysterious ability to make something out of nothing. It is the ability to connect things that were far apart, draw a different question from a familiar object, and give form to a small difference that no one had considered important.

Difference begins with observation, but it is completed as value.

The Questions That Remain After Automation

In the industrial age, the ability to produce products of consistent quality through repetition was important. In the digital age, the ability to find, connect, and distribute information became important. In the AI era, much of execution and expression is being automated. The cost of producing an output will fall, and the average level of finish will rise.

Human competitiveness is therefore likely to move toward abilities such as these: sensing differences that do not yet have names; judging why those differences matter; and giving them a form that other people can experience.

Automation lowers the cost of repetition. Detecting a difference finds a direction. Generating a difference opens a new possibility. That possibility becomes value only when it reaches another person’s life.

So the important question in the AI era should not stop at “How much work have we automated?” It should ask: With the time and ability automation has secured for us, what difference that did not exist before have we created in the world?

Deleuze’s Difference and Repetition can seem like a philosophy that prepared this question long ago. The future will not open only to those who perform the most repetitions. It will open more widely to those who discover differences that were not yet visible within repetition and turn them into value that a community needs.

---

References

[1] Erik Brynjolfsson, Danielle Li, Lindsey R. Raymond, “Generative AI at Work,” The Quarterly Journal of Economics 140(2), 2025, pp. 889–942.

[2] Anil R. Doshi, Oliver P. Hauser, “Generative AI Enhances Individual Creativity but Reduces the Collective Diversity of Novel Content,” Science Advances 10(28), 2024.

[3] Fabrizio Dell’Acqua et al., “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality,” Organization Science 37(2), 2026, pp. 403–423.

[4] Gilles Deleuze, Difference and Repetition, Columbia University Press; Daniel W. Smith, John Protevi, “Gilles Deleuze,” Stanford Encyclopedia of Philosophy.