My LinkedIn feed is full of discussion about AI as a teaching tool, or students’ use of it in schools. I see far less about what it means for professionals in education.

This is the first in a short series of articles on what I believe it actually means to use AI well in a professional context. I’ll start with a question: what gives away an AI-generated response for you?

Disclosure 1: I used AI as a thinking partner in developing this piece. I will discuss what this means in practice, and why it matters, in a future instalment in this series.

Disclosure 2: As a proud, long-time user of the em dash, I have been dismayed by its relegation to ‘AI indicator’ status and have tried — with partial success — to rein myself in for these articles.

· · ·

I had a conversation with a colleague recently about something we’d both started noticing: the growing ease with which you can identify an AI-generated response in the wild. Not because AI writes badly, but because it doesn’t write like the person sending it. In my experience, this is especially true of bullet-pointed feedback on a shared document or presentation.

We’d both been in situations where we received a response to something and, almost immediately, thought: that’s not them. The register was off. The structure was too clean. The points were too generic. We’d gone to these people specifically for their views, and it didn’t feel like we were getting them. What we were reading was perfectly fine. But perfectly anonymous.

I don’t think this happens because people are lazy, per se. Rather, it comes from two understandable instincts. The first is a kind of intellectual anxiety: a fear of missing something you shouldn’t, or of not expressing yourself as well as AI might. The second is a burgeoning faith that AI is simply better at this than we are. That our job is merely to prompt it as best we can, and resign ourselves to dispatching its outputs verbatim.

Both instincts are wrong, at least for the time being, and both carry a cost.

At best, a reply that doesn’t sound like you suggests you were short on time, or trying too hard to be efficient, and didn’t think the communication warranted a personally crafted response. At worst, the recipient may be left with the impression that you don’t have a view of your own: that you had to rely on AI to do the thinking for you. In professional contexts where your judgement and expertise are precisely what you’re being paid for, that is a significant thing to signal — even inadvertently.

There is a more subtle cost, though, and in the long run a more serious one. When you outsource a response entirely, you don’t just risk how others see you. You erode something in yourself. Forming and expressing a view, even on a small thing, is how you keep your thinking sharp. Skip it often enough and cognitive atrophy sets in (think Dr Kawashima and his Nintendo DS brain training; if you’re of a significantly different generation to me, Google it. Or TikTok it. Or whatever you do).

And then there is the problem of the echo chamber. If the responses circulating in a conversation have all been filtered through the same tools with the same defaults, what purports to be a field of perspectives from experienced professionals is, in reality, anything but. The divergent thinking that would have done the most important work gets smoothed away before anyone has the chance to interact with it.

People still make intellectual leaps that AI struggles to replicate. The kind that draw on lived experience, creative instinct, and the particular way an individual mind makes connections. An AI can synthesise and challenge and refine. I am not convinced it can have an idea that surprises itself.

Used well, AI is a sparring partner, not a ghostwriter. It can challenge your phrasing, push back on your reasoning, and — if you let it (that’s the fourth em dash. I know. Here comes the fifth) — force you to dissect your own position and understand where it actually came from. That last part is more important than it might seem. A view you can’t explain the origin of is a view you probably can’t defend.

So, I have started to investigate what I can do to make AI work for me without falling into these traps. In the next instalment, I’ll write about how I am approaching this. Spoiler: it involves building something that gives AI enough context about the way I naturally articulate my thinking that it can do that sparring job properly. Thus far, it has been a more revealing exercise than I expected.

P.S. Comments very welcome, including your answer to the opening question. Bonus points if you can spot where a sixth (and seventh…?) em dash would have sharpened a sentence.