Five advantages of AI, depending on who's trying to sell it to you

Today we talk about the twin of the listicle we saw in 0081. That one listed dangers; this one lists advantages. The piece we're going to take apart always has the same title —"Five advantages of AI in your company," "5 reasons to use ChatGPT," "Five benefits of artificial intelligence"— and the same internal structure. Productivity, cost savings, personalization, automation, competitive advantage. Five identical bullets across fifteen different corporate blogs. Why always five and always the same? Because the advantages listicle has no informative function: it has a legitimizing function. Turning a political choice into a neutral list of obvious benefits. My take is biased because I wrote one of these in a former life, for money, and it still embarrasses me. Form your own.

I've kept a notebook since 2022 parallel to the dangers one. I jot down every advantages listicle that passes through my feed. I'm at forty-one. Thirty-two have exactly five points. Five have seven. Four have ten. None has six, three, or nine. The distribution isn't chance; the ecosystem has decided that five is the optimal figure for this genre.

The standard structure, told honestly

If you add up the fifteen most-read pieces in the Spanish SERP for "advantages of artificial intelligence" in 2024-2025, you find a common structure with barely small variations. Point one, productivity: AI lets you do more in less time. Point two, cost savings: AI reduces operating spend. Point three, personalization: AI adapts the experience to each user. Point four, automation: AI frees the worker from repetitive tasks. Point five, competitive advantage: AI lets the company get ahead of its competitors.

Five bullets. Almost always in that order. Almost always with the same language. The text accompanying each bullet rarely tops 100 words and rarely includes verifiable sources. There are generic examples —"for instance, a company can automate customer service with a chatbot"— but no concrete cases with names, figures and verification. The piece, all told, runs between 700 and 1,200 words.

Where does this identical structure come from? From the press releases of the big makers and from the corporate blogs of the consultancies that sell integration. The trail is direct. When you read the blog of IBM, of Microsoft, of Salesforce, of Deloitte, of Accenture, of KPMG, you find the same scheme in pieces signed by marketing teams. The external copywriters of SEO blogs, the community managers of small businesses and the blogs of general outlets reproduce that scheme because it's the most accessible source and because the result works in search engines.

What each bullet hides when you look closely

Here comes the uncomfortable exercise. Let's take each of the five points and ask the question the listicle never asks: an advantage for whom, at the cost of what?

Productivity. When a listicle says "AI increases productivity," the figure that usually accompanies it is on the order of 20% to 40% in specific tasks. It's true. What isn't said is for whom. If the worker does the same in less time and the company now asks them for more tasks in the same hours, the company captures the advantage entirely through work intensification. If the worker does the same in less time and the company reduces their hours without reducing pay, the advantage is shared. McKinsey's figures in State of AI 2025 show that the first scenario is far more common than the second. The "productivity" of the listicle is, in most cases, intensification.

Cost savings. When a listicle says "AI cuts costs," the typical figure is on the order of 10% to 25% in operating expenses. The operational question: where does that saving come from? Almost always from headcount. "Automating customer service" means fewer call-center agents. "Automating administrative processes" means fewer administrative staff. "Automatic report generation" means fewer junior analysts. The saving is real for the company and it's real unemployment for the workers no longer needed. Saying "cost savings" without naming who pays for that saving is telling half the story.

Personalization. When a listicle says "AI personalizes the experience," the mechanics are always the same. The AI takes the user's interaction data and adjusts the service. The advantage is presented as the user's —"experience adapted to you"— but the raw material needed to produce it is the user's own data, which becomes a permanent asset of the company that collects it. The asymmetry is brutal: I get a marginally better recommendation; the company gets a permanent behavioral profile it can use for training, for ad targeting, or for sale to third parties depending on its privacy policy. The personalization of the listicle is input capture.

Automation. When a listicle says "AI frees the worker from repetitive tasks," the framing is always positive. What isn't said: many of those "repetitive tasks" are precisely the ones a junior worker uses to learn the trade. When the drafting of meeting minutes is automated, the junior doesn't draft them and doesn't learn to synthesize. When the production of basic reports is automated, the junior doesn't do them and doesn't learn the structure. The consequence is generational. The current junior generation learns less of the operational trade because the learning tasks have been automated. Economists like David Autor of MIT already discuss this in his more recent papers on AI and the Labor Market (2024-2025).

Competitive advantage. When a listicle says "AI gives competitive advantage," the framing is misleading by construction. If all the companies in the sector adopt the same technology at the same pace —which is what's happening—, the relative advantage is neutralized. What remains is a higher structural step: the same level of competition but with greater technological dependence and higher maintenance cost. The "competitive advantage" the listicle sells flattens out over three to five years as everyone competes on similar terms. The only sustainable advantage is that of whoever arrived first and stayed in the supplier's position —the hyperscalers—, not that of whoever adopted as a client.

The shared language, the corporate metalanguage

There's a linguistic phenomenon worth naming because it runs through all five points. The advantages listicle uses a vocabulary the trained ear recognizes immediately. Terms like "scalable," "agility," "added value," "empowerment," "synergy," "digital transformation," "technological adoption," "disruptive solutions." Each of these terms has a recognizable origin in consultancies or corporate marketing of the nineties and two-thousands.

The characteristic of these terms is that they seem to say something concrete and don't. If you try to define "scalable" precisely in the context of a small company, you discover the term refers to the capacity to grow without a linear proportion in costs, but the listicle never quantifies that proportion. If you try to define "agility" operationally, you discover the term means "the capacity to respond quickly to changes," but the listicle doesn't specify what changes or what speed.

This corporate metalanguage performs a specific function. It makes the text sound technical and professional without forcing the author to commit to anything verifiable. It's what Harry Frankfurt called, in his short essay On Bullshit (Princeton University Press, 2005), "bullshit" in the technical sense: discourse produced with no regard for the truth, not even for the lie; produced to occupy the communicative space with the appearance of content.

The average listicle about the advantages of AI is, in Frankfurt's terms, a high concentration of professional bullshit. It doesn't actively lie —the individual points are technically defensible— but it's designed not to inform, but to legitimize a decision the reader will make for other reasons.

The concrete economy of the advantages listicle

The operational question is why so many are produced. The answer is economic and worth seeing clearly.

The advantages listicles are optimal SEO content for B2B lead capture. A consultancy that sells AI integration services makes money every time a company hires it to implement AI in its processes. The sales process begins, in most cases, with the buyer searching for information on Google. If the first page of results is dominated by listicles presenting AI as advantageous without nuance, the buyer arrives at the sales funnel with a favorable prejudice. The consultancy closes the contract more easily.

The listicle sells better than analysis. Honest analysis —"it depends on your sector, your size, your client, your internal technical capacity; here are the risks; here's the list of questions you should ask yourself before adopting"— doesn't produce immediate leads. It produces doubts. The consultancy prefers manufactured certainty.

The result is an ecosystem where the most visible content is the most superficial, and the most superficial is the most aligned with the interests of whoever's selling. That correlation isn't accidental. It's the business model of B2B content marketing applied to the AI sector.

The question the listicle never answers

There's a simple question the format doesn't allow you to answer, and that's why the format doesn't include it. The question is: an advantage, for whom? Artificial intelligence isn't a neutral phenomenon that produces uniformly distributed advantages. It's a set of technologies whose benefits and costs are distributed asymmetrically among different actors.

For the supplier hyperscalers —Microsoft, Amazon, Google, OpenAI, Anthropic— AI is a clear, quantifiable advantage. Their cloud business grows, their margins rise, they capture capital.

For the large companies that adopt AI strategically, AI can be a temporary advantage if they arrive first, neutral if they arrive on pace, or costly if they arrive late and badly.

For the small companies that adopt AI out of fashion, AI is usually a cost with a diffuse short-term return. McKinsey documents that 5.5% of organizations report a clear financial return; the rest are paying licenses and implementation with benefits not yet aggregable to EBIT.

For the workers whose tasks are automated, AI is a direct threat to the job or to the job's conditions.

For the workers whose tasks are complemented —not replaced—, AI can be a tool that increases personal productivity and, depending on how that gain is distributed, can be a benefit or a work intensification.

For the citizens who generate data by consuming AI services, AI is an asymmetric exchange where you get a free or cheap service and hand over a training corpus.

That diversity of positions —which is the reality of the sector— doesn't fit in a five-bullet listicle. It fits in a structured analysis of several thousand words like the one you're reading. It fits in books like Power and Progress by Acemoglu and Johnson or Atlas of AI by Crawford. It doesn't fit in the format the media and commercial ecosystem rewards.

The political question

This is personal opinion, but the economic and media data support it. The advantages listicle of AI is the contemporary version of an old genre: the sales pamphlet disguised as information. There's no individual malice in whoever produces it. There's an economic system that rewards its production and an informational ecosystem that distributes it.

The political consequence is serious. Decisions to adopt AI in mid-sized Spanish companies are being made, in large part, on the basis of information that comes from the listicles of the consultancies that will later bill for the integration. The conflict of interest is structural and rarely named. The managers of small and mid-sized companies, the board members of cooperatives, the directors of schools, the chiefs of staff of regional administrations: many of them reach the decision to adopt AI with their heads full of listicles and empty of rigorous analysis.

What I would ask —and again with no realistic hope of changing it in the short term— is that any advantages listicle of AI be obligatorily accompanied by three pieces of information. First, who funds the piece (declared advertising, affiliation, source press release). Second, what risks the piece doesn't cover. Third, what questions the reader should ask before adopting. Those three pieces wouldn't make the listicle less sellable, they'd make it honest. But precisely because it'd be honest, it wouldn't work as well as an instrument of capture, which explains why nobody is going to add them voluntarily.

In the meantime, the individual reader can at least recognize the genre. When you see a listicle of five advantages of AI, ask yourself three things: who publishes this article and what do they sell? What's missing from the piece? For whom is each bullet an advantage? If you answer those three questions, you're already reading the text with judgment. What the listicle doesn't let you read is what the listicle doesn't want read.

The concrete figure that closes it. According to McKinsey's report The State of AI in 2025: Agents, innovation, and transformation (November 2025), only 5.5% of the organizations surveyed report a clear financial return on their AI investment, while over 80% see no tangible impact on corporate EBIT. That figure —5.5%— is the quantitative translation of the question "an advantage, for whom?" For 5.5%, yes. For 94.5%, not yet. The average listicle mentions neither of those two figures, and it's precisely that omission that defines the format.

Definitions

Advantages listicle: a web genre structured as a numbered list of benefits of a product, technology or service. The positive version of the dangers listicle seen in 0081. It performs a legitimizing rather than informative function.

Lead generation: a commercial process by which an anonymous visitor to a web page voluntarily provides their contact details in exchange for apparently free content. It's the main business model of B2B consultancies on the internet.

Corporate metalanguage: a specific vocabulary —scalable, agility, synergy, digital transformation— that seems technical but doesn't commit the speaker to verifiable claims. It serves to produce professional discourse without specific content.

Bullshit (technical sense): discourse produced with no regard for the truth, not even for the lie, but for the effect of the discourse on the recipient. A concept developed by Harry Frankfurt in On Bullshit (2005).

References

Harry G. Frankfurt, On Bullshit (Princeton University Press, 2005). A philosophical framework for understanding corporate discourse with no commitment to truth.

McKinsey & Company, The State of AI in 2025: Agents, innovation, and transformation (November 2025). Figures on the real return of business AI investments.

Stanford HAI, Artificial Intelligence Index Report 2026 (April 2025). Section on business adoption and public conversation.

Daron Acemoglu & Simon Johnson, Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity (PublicAffairs, 2023). An economic framework on how the benefits of a technology are distributed and why the direction isn't neutral.

David Autor, Applying AI to Rebuild Middle Class Jobs (NBER Working Paper, 2024). Recent research on how AI affects low- and mid-skilled work.

Edward S. Herman & Noam Chomsky, Manufacturing Consent: The Political Economy of the Mass Media (Pantheon, 1988). The classic text on the filters that act on information in media with economic dependence.

Reuters Institute for the Study of Journalism, Digital News Report 2025 (Oxford University, June 2025). Data on the consumption and production of SEO content in digital media.

Going deeper

Ryan Holiday, Trust Me, I'm Lying (Portfolio, 2012; revised edition 2017). An insider's manual on how content is designed to capture attention and conversion.

Kate Crawford, Atlas of AI (Yale University Press, 2021). A material and political analysis of AI that the advantages listicle deliberately hides.

Shoshana Zuboff, The Age of Surveillance Capitalism (PublicAffairs, 2019). A framework on how data capture via "personalization" reconfigures the digital economy.

Ana Trujillo Cabezas, Cómo se escribe el SEO: manual práctico del periodismo digital (sector manuals, various years). Operational documentation of the trade of the SEO content writer in Spanish.

You may also like

• The 10 dangers of AI and pigs fly
• Ten reasons never to read another listicle about AI in your life
• OpenAI's press release isn't journalism
• Inflated productivity

Elsewhere

• AI Now Institute
• OECD.AI Policy Observatory