The most concerning result of AI in professional communication is not bad writing. It is writing that is technically clean and intellectually absent.

There is a distinction worth making between two kinds of AI assisted professional communication. The first is work that AI has helped make better: clearer, better structured, more precisely worded. The human thinking is present. The AI has sharpened the expression of it. This is useful and produces genuinely better communication.

The second is work that AI has replaced. The structure is there. The vocabulary is appropriate. The argument flows logically from paragraph to paragraph. But there is no actual point of view in it. No specific observation that could only have come from this person’s experience. No position that required courage to take. It is, in the phrase I use with my students, hollow competence: perfect grammar, zero understanding.

What Hollow Competence Looks Like

A student submits three consecutive assignments that are technically clean. Good structure, appropriate vocabulary, logical progression. But when asked to explain the central argument in a tutorial without looking at the paper, she cannot. She knows the topic. The ideas exist somewhere in her understanding. But the writing was not the place where she worked out what she believed, because the writing was not hers.

The first version came from Seoyeon, a 27 year old brand marketer at a skincare company in Seongsu who writes a quarterly review in English for the regional office in Singapore. She had pasted her notes into a chatbot and asked for a professional summary. Nothing in it was wrong. When her team lead asked in the meeting what she would actually do about the price cuts, she had no answer, because the paragraph had never asked her to decide anything. The second version took her forty minutes and two arguments with herself. It is shorter and less polished, and it is the only one of the two that could only have been written by someone who works there.

This is not a niche problem in Korea. The Ministry of Science and ICT reported in its 2025 Internet Usage Survey (인터넷이용실태조사) that 44.5% of Koreans have used generative AI, up 11.2 percentage points on the year before. In my experience, a large share of that use at work is exactly this kind of writing, and the habit has spread much faster than the habit of checking what comes back.

This is the gap the world is noticing. The Stanford AI Index 2026 finds optimism and nervousness rising together: 59% say AI products bring more benefits than drawbacks, 52% say they make them nervous (Chapter 9, Figure 9.1.1). Perfect grammar with zero understanding is exactly the output that feeds the second number.

A communications team produces a campaign brief that passes every internal review. The proposition is well worded. When the agency comes back with a creative that misses the point completely, the team struggles to explain specifically what is wrong, because they never actually argued about what the brand believes. The brief documented a consensus that was never really reached.

Research published in the Journal of Marketing and Social Research in 2025 found that AI generated content significantly reduces perceived authenticity and brand trust compared to content made by humans, and that disclosing AI authorship makes the trust penalty worse, not better. Audiences are detecting not the tool but the absence of a genuine perspective.

The Competence Hierarchy

The most useful frame separates three levels of engagement with any tool. Owning it: having the subscription, knowing it exists. Operating it: knowing how to prompt effectively and integrating it into the workflow. This is what most AI literacy programmes teach. Thinking with it: using AI to extend and test your own reasoning rather than to replace it. Hollow competence is what happens when someone jumps from the base level to the output of the top level without doing the cognitive work in between.

The World Economic Forum’s Future of Jobs Report consistently places analytical thinking and creative thinking above technological literacy in its ranking of critical professional skills. The tools are only valuable to the degree that genuine thinking sits above them.

The Test

The test I use in my technology and AI course is simple: can you explain your reasoning without a screen? Not read it back. Explain it. Can you say, in your own words, what you believe and why, and defend it when someone pushes back? If the answer is no, you have not developed the competence. You have borrowed the appearance of it. The WiFi goes down. The subscription lapses. What remains in those moments is what you actually built.

Running the Test on a Real Draft

The test only works if you run it before someone else does. Here is the routine I gave Seoyeon. It takes about as long as the chatbot saved her in the first place.

Step two is the one people skip, and it is the one that matters. If you underline the facts in a hollow draft, you usually find that every one of them could have come from a search engine. That is the signal. A 2025 study from the MIT Media Lab followed 54 participants writing essays with an LLM, with a search engine, or with no tools at all. Ownership of the essays, as participants reported it, was lowest in the LLM group, and the researchers note that “LLM users also struggled to accurately quote their own work.” If you cannot quote it, you cannot defend it.

Why Hollow Competence Is Easy to Miss in a Korean Office

Three things make the problem harder to spot here. Reports travel up a hierarchy, so the person reading the draft is often senior and busy, and reads for form before substance. Reports in English are often read by people working in their second language, who in my experience are slower to question fluent prose. And the culture of the polished 보고서 rewards documents that look finished. A clean AI paragraph fits all three, which is why it can travel all the way to a director before anyone asks a hard question.

Run these three on your own draft before the meeting, not during it. If your honest answers sound like the left column, the paragraph is not ready, however clean the grammar is. The University of Melbourne and KPMG global study of more than 48,000 people across 47 countries found that many people rely on AI output without evaluating accuracy (66%) and are making mistakes in their work due to AI (56%). The mistakes are the visible cost. The left column is the invisible one.

If you want a second reader before the human one, paste your draft into Copy Check and see whether the specific lines survive. The Offload Line helps you decide which parts of a task you can hand to a tool and which have to stay with you. For the habits underneath all of this, read The Prompt Is Not the Skill, why the first draft matters most, how to write like a human in 2026 and why AI literacy can no longer be optional.

→  This is one of the topics I speak on in classrooms and also for corporate audiences: the distinction between AI assisted competence and AI substituted competence. The Work With Us page has more on speaking engagements and workshop formats.

If you want practical tools to sharpen how you communicate professionally, the communication tools on this site are a useful starting point.

---