I was on the radio for World Hepatitis Day. The producer had sent the questions in advance, and I had worked through all of them: transmission, testing, who should be vaccinated. By the time we went on air, I was comfortable. Then a listener called in and asked about mental health. Not a harder question about hepatitis. A different subject altogether. There was no version of that moment where I could say "let me come back to you on that," no draft to revise, no minute to think. The microphone was open, and the answer had to come from me. It went fine. But something occurred to me in those few seconds that I have been turning over ever since. That caller did not ring in because of the topic of the programme. He rang in because someone was there who might help him with what was actually on his mind. He was not responding to a subject. He was responding to a person. And I know a lot of young professionals who now produce excellent written work. Their applications are clean. Their proposals are structured. Their concept notes open with a clear problem statement. Much of it is drafted with AI, and it is good. Almost none of them are putting the same effort into learning to speak. What actually changed For a long time, writing well worked as a filter. It was a rough proxy for how carefully someone thought and how seriously they took the opportunity. Never a perfect signal, but it sorted people. Those who could write moved ahead of those who could not. That filter has mostly dissolved. When competent writing is available to everyone at no cost, it stops telling you anything about the person who submitted it. So the weight moves elsewhere. It moves to the parts of professional life that happen live. The room did not change Consider what a working week actually asks of you. Someone has to present the proposal to the funder and answer the question about sustainability that isn't in the document. Someone has to sit in front of an interview panel and account for a two-year gap in their CV. Someone has to stand in a departmental meeting and explain why a new protocol is worth the disruption, to colleagues who have watched protocols come and go. In clinical work, the exposure is sharper. You explain a diagnosis to a patient who is frightened and not fully following you. You take consent in a way that is real rather than procedural. You tell a family the outcome was not what anyone hoped. You raise a concern about a senior colleague's decision, in the moment, out loud. None of that has been automated, and I don't expect it to be. Every one of those moments needs a person who is present, thinking in real time, and accountable for what comes out of their mouth. What people are judging When someone listens to you speak, they are not grading your grammar. They are running a much older assessment. Do you understand this, or did you memorise it? Are you the same person now that you were on paper? What happens to you under pressure? Can you say "I don't know" without falling apart, and do you know what you would do next? That last one carries more weight than people realise. A confident wrong answer damages trust far more than an honest gap. Patients sense it. Panels sense it. Colleagues sense it fastest of all. None of these judgments can be made from a document, because a document is edited, unhurried, and revisable. Speech is none of those things. You get one pass, in front of people, with no chance to go back and fix the sentence you have already said. That is exactly why it carries information a file cannot. What this means in practice Four things, in rough order of usefulness. Read your written work aloud before you send it. If you cannot say it in your own voice, you will not be able to defend it in a room. It is the fastest way to find out whether you actually understand what you submitted. Prepare for the question that isn't on the list. Most people rehearse the part they control and neglect the part they don't. I had prepared for hepatitis; the question was about mental health. Write down the three questions you hope nobody asks — and then accept that the real one may come from somewhere you didn't think to look. Practise saying "I don't know." Not as an apology, but as a complete professional answer: what you don't know, why, and what you would do to find out. Rehearse it until it sounds calm, because you will need it when you are not. Take the small speaking opportunities. The CME slot nobody wants. Introducing a guest. Presenting a case at the morning meeting. Chairing a staff meeting. These are low-stakes repetitions, and there is no substitute for repetition. Nobody becomes comfortable in front of a hard panel by reading about it. The part you cannot delegate I am not arguing that anyone should stop using these tools. I use them. They save real time, and refusing them on principle is not a strategy. But it is worth being clear about what they do for you and what they leave on your desk. They can produce the document. They cannot produce you. Writing was never the hard part of professional life. It was the part that used to be scarce. What was always hard, and remains hard, is standing in front of people who have every reason to doubt you and giving them a reason not to. That is still yours to do.