AI will not create a competitive advantage

Over the past year, enthusiasm for artificial intelligence seems to have moved beyond the discovery phase and entered the industrialization phase. Every week, new tools appear, use cases are validated, and the discourse aligns with a quasi-single objective: automate, accelerate, optimize. But this leads me to ask myself a question: if everyone is using the same tools to do the same thing in the same fashion, where has differentiation gone? Moreover, if AI is supposed to produce excellence, what does the notion of excellence mean when it becomes accessible to everyone?

This question is far from new. Nicholas Carr asked it in 2004, in relation to information technology, in a book that I loved at the time: Does IT Matter? Information Technology and the Corrosion of Competitive Advantage. It is back in the spotlight today because AI, by making formalized knowledge imitable, threatens to transform competitive advantages into industry standards: when excellence becomes the norm, it becomes commonplace and is therefore no longer excellence. It is this mechanism of standardization, and its organizational, human, and strategic consequences, that we will discuss in this article.

In short:

• The industrialization of AI is leading to a standardization of practices that reduces differentiation between businesses, making formalized knowledge easily imitable.
• Like information technology in the 2000s, AI is becoming a commodity: it improves performance but no longer constitutes a sustainable competitive advantage.
• Excessive automation can impoverish the customer experience (e.g., Klarna) or reproduce biases (e.g., Amazon), stripping interactions of their human and contextual richness.
• Key skills such as intuition, judgment, and taste cannot be automated and must be protected as sources of strategic differentiation.
• To maintain an advantage, businesses must create environments for learning and embodying knowledge, where AI supports but does not replace human development.

The illusion of differentiation through technology

In the 2000s, Nicholas Carr drew the ire of the tech establishment when he published Does IT Matter?, in which he argued that information technology, by becoming ubiquitous, was no longer a lever for differentiation from the competition. His reasoning was simple: information technology, once rare and complex, was now accessible to all, standardized, industrialized, and implemented by the same firms, with the same tools, using the same methods. IT had become a commodity, an infrastructure like any other, comparable to electricity or running water, and what had once been a competitive advantage was now an obligation of functional compliance.

Twenty-one years later, artificial intelligence is following exactly the same path. With the initial euphoria gone, businesses are rushing to equip themselves with the same tools, the same pre-trained models, and the same consultants. They are all seeking to automate what can be formalized, document tasks, and capture processes in prompts and decision trees, but this race for mass deployment, without strategic thinking about differentiating use, does not produce an advantage. On the contrary, it creates a kind of uniformity, and when everyone is doing better, no one is doing better than the others.

When it comes to technology, the distinction between what creates an advantage and what is simply the price of entry into a market has never been so thin. AI, in its current uses, is becoming the new ERP: it keeps the business running, but it no longer sets it apart (A software that helps to streamline processes ? Run away !). Worse still, as we will see, it can easily lead to a leveling down of practices.

Klarna: from optimization to disillusionment

The example of Klarna is quite symptomatic of this problem. The Swedish startup announced the elimination of 700 customer service positions, replaced by chatbots, and congratulated itself on having achieved peak efficiency, which led to more than 2,000 job cuts. The scripts were perfect, the decision trees impeccable, and, on paper at least, everything was running smoothly. But in reality, customers were leaving.

Faced with the deterioration of customer relations, Klarna suddenly backtracked, rehired staff, and reintroduced humans where AI had replaced everything. Because while the scripts said “I understand your frustration,” customers understood very well that they were not being understood. Automated speech had lost the reassuring and human aspect of the exchange. By automating the relationship, the business had, in a way, sacrificed the intention.

This story is perfectly told by Jean-Paul Paoli (When AI turns your secret sauce into ketchup), who shows us how the automation of customer service, by emptying the exchange of its humanity, has generated a uniform and not very engaged experience. What the business automated, it made imitable, and in doing so, it crystallized its customer relationship into a standard, replicable, off-the-shelf form. By believing it was transforming its excellence into a system, it emptied it of its substance.

The words of its CEO at the time of the backpedaling are eloquent, to say the least: “We just had an epiphany: in a world of AI nothing will be as valuable as humans! Ok you can laugh at us for realizing it so late, but we are going to kick off work to allow Klarna to become the best at offering a human to speak to!!!”

Amazon’s mistake: when AI learns from past biases

The case of Amazon is equally instructive. Between 2014 and 2017, the company developed a recruitment algorithm based on ten years of internal data with the aim of identifying the best candidates. The result: the algorithm learned to discriminate, penalizing female applicants, devaluing resumes from women’s colleges, and reinforcing past biases by embedding them in code.

The case was revealed in 2018 by Reuters (Insight – Amazon scraps secret AI recruiting tool that showed bias against women): in attempting to automate human judgment, Amazon unintentionally reproduced its own historical prejudices. What this experience shows is not only the danger of feeding AI with biased data, but also the illusion that human discernment can be translated into formal rules without losing what makes it so valuable: nuance, context, and responsibility.

The business thought it was securing its ability to recruit by systematizing it, but in reality, it undermined its ability to judge. Formal intelligence, the kind that can be articulated, encoded, and simulated, is always one step behind embodied intelligence, which feels, adjusts, and takes the risk of sometimes stepping out of line based on intuition.

A true competitive advantage cannot be automated

Here, we need to think in terms of the opposite of the prevailing discourse. Today, AI makes it possible to automate everything that is documented, explicit, and reproducible, which frees up time, reduces costs, and improves productivity on tasks of low cognitive value. But what AI makes easy, it also makes commonplace. The easier it is to transfer knowledge to a machine, the less distinctive it becomes in a competitive context.

Conversely, what remains difficult to simulate becomes precisely what deserves to be protected: intuition, taste, judgment. It is these forms of embodied knowledge that constitute the real competitive advantage, not because they are magical or mysterious, but because they are the fruit of history, culture, and experience. Intuition cannot be decreed, it is forged; taste cannot be simulated, it is experienced; and judgment cannot be coded, it is exercised.

As Paoli points out, Louis Vuitton does not have an algorithm for luxury, just as Apple cannot document “delight” or a good manager cannot make decisions based on a binary tree. All these forms of knowledge are based on thousands of integrated micro-experiences, on an ability to feel before knowing, to guess before modeling, and it is these that distinguish one business from another. It is these that AI, by its very nature, cannot produce and that its massive deployment paradoxically risks eradicating.

The AI paradox: less practice, fewer future experts

This is where the pitfall of AI automation lies: in seeking immediate efficiency gains, we are destroying the conditions for the emergence of tomorrow’s skills. The surgeon who operates alone with a robot no longer passes on their knowledge, the junior who corrects the outputs of an LLM no longer builds professional intuition, and the project manager who has Copilot proofread their strategic recommendations no longer faces doubt.

All of this has been described very well by Matt Beane, notably in The Skill Code, where he observes that automation tends to interrupt traditional learning chains: novices make fewer and fewer mistakes, but they also learn less and less, because they are no longer sufficiently integrated into the processes where, before, their judgment was refined. By eliminating “tedious” tasks, we also eliminate learning opportunities, and in believing we are saving time, we lose out on training. Optimization comes at a price: disembodiment and loss of judgment.

The businesses that will succeed tomorrow will not be those with the best tools, but those that have managed to maintain learning paths, moments of confrontation with reality, and spaces for slow training in judgment.

Protecting your competitive advantage with spaces for embodiment

It is therefore imperative to think differently about the integration of AI, not as a means of eliminating human work but as a lever for enhancing it. AI should be reserved for tasks that are articulable, explicit, and repetitive. Everything else like human interaction, creativity or uncertainty should be seen as strategic, a kind of space for differentiation.

This means preserving certain ambiguities and gray areas, not documenting everything, and not turning every corporate culture into a playbook. It also means forcing embodied experiences: sending product managers into the field, having data scientists listen to customer calls, creating workshops where juniors don’t just validate prompts but construct their own thinking.

It also means recreating pairs, human-AI tandems, where the machine executes, but the mentor explains, corrects, and conveys the invisible: when to break the rules, weak signals, implicit context. Finally, it means seeing certain things that are considered inefficient or unproductive (learning, repetition, feedback) not as costs, but as investments in judgment.

Bottom Line

In the rush towards artificial intelligence, it is tempting to believe that victory will belong to those who move fastest, who automate the most, who document everything in order to model everything. But recent history, from ERPs to chatbots, teaches us that technology is only as good as the use we make of it, and the value it creates always depends on what escapes it.

AI can certainly become a performance lever, but it will never become a lever for sustainable differentiation if it merely imitates, summarizes, and responds. What distinguishes businesses, what constitutes their corporate culture, their DNA, their decision-making process, cannot be coded but is built slowly, through experience, doubt, and confrontation with reality. And this is transmitted not through automation, but through learning.

At a time when everything that can be formalized is becoming reproducible, it is vital to preserve what cannot be formalized, such as intuition, judgment, and the ability to understand before explaining. The only businesses that will be able to create an advantage that AI cannot steal from them are those that know how to design environments conducive to the embodiment, transmission, and development of this knowledge. Not despite AI, but with it, provided that it is given the right place.

To answer your questions

Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI).

Related articles:

Will AI replace juniors? The false debate that’s only the tip of the iceberg AI and talent management: a dynamic approach that promotes internal mobility Will AI change the way performance is measured and rewarded? Automation is done with people, not against them. Let’s stop being nAIve with AI in the workplace HR/IT and the reality of working at Moderna: the unspoken truths of a reorganization