I’ve been watching a Netflix series called Agent Kim Reactivated – a lively, ongoing K-drama along similar(ish) lines to the classic film franchise Taken, but less… earnest. One particular action sequence caught my attention: it was slick, cinematic and surprisingly ambitious for a three-minute flashback sequence. Explosions, car chases, gunfights – telling a crucial bit of the backstory of the lead character. I remember thinking it looked really, really good. I had a little marvel at the production values.
Two days later – via an Instagram post (because, like a lot of people my age, I like to get my entertainment news a week or so late rather than engage with TikTok) – I learned the entire sequence had been created using AI. The producers had been very open about it; they describe it as the first Korean TV drama to use AI to generate a complete story-critical action sequence rather than just individual visual effects shots. They argued it allowed them to tell an important part of the story without the huge cost of overseas filming, practical effects and extensive post-production.
What surprised me wasn’t that AI could produce something of that quality. We’ve all seen enough impressive demos over the last couple of years to know that’s possible, and I’d assumed there was some heavy use of CGI. It was that I hadn’t realised it was straight-up AI-generated.
It can’t just be me
I suspect I’m not alone. We all like to think we can spot AI – social media is full of people confidently announcing that an image was “obviously AI”, or an email was “clearly ChatGPT”. Sometimes they’re right. Increasingly, I suspect they’re congratulating themselves after the fact. Once you know something was AI-generated, it’s remarkably easy to identify all the clues. Before you know, it’s much harder. It’s like those optical illusions that become impossible not to see once someone points them out.
And it’s not that the Agent Kim sequence was trying to deceive me. Nor because I was watching particularly closely or not closely enough. It simply fitted, and it was really good. It looked like it belonged there. I’m generally in awe of Korean TV and film production values, and my brain accepted it without a second thought.
That feels like quite an important moment, for me if no one else. I work in AI and have an actual degree in Film Studies.
The question has changed
For the past couple of years, we’ve spent an enormous amount of time asking whether AI is good enough for various tasks. Every new model has prompted another round of comparisons. Can it draw left-handed people yet? Does the voice sound natural? We’ve become expert at looking for the seams. Or so we think.
The interesting thing is no longer whether AI can produce convincing work. It’s how often we stop noticing that it has. Because I don’t think this is just happening in television. I think we’ve reached the point where we’re exposed to AI-generated work every day without giving most of it a second thought. Not because we’re gullible, but because “good enough” increasingly doesn’t announce itself.
It’s happening in emails, reports, presentations, meeting notes, marketing copy, images, translations… The average standard of so much everyday work has risen remarkably quickly. Not because AI is making everything brilliant, but because it’s making an extraordinary amount of work perfectly competent.
AI isn’t replacing excellence. It’s replacing obvious mediocrity.
Twenty years ago, polish was a skill in its own right. Producing a good paper, a professional-looking slide deck or a decent promotional image took time and experience. Even getting your thoughts into well-written prose wasn’t straightforward for many people. Those things carried value because they were relatively scarce.
Today, polish is pretty abundant, and that’s mostly a good thing. AI helps people communicate ideas they might otherwise struggle to express. It lowers barriers and helps small organisations that can’t afford top-end designers or copywriters and used to have to settle for some pretty basic stuff. It allows independent film-makers to attempt scenes that would previously have blown their budget.
But every technological shift changes what is scarce. If everyone can produce work that looks competent, competence itself becomes less distinctive. AI doesn’t make everyone brilliant; it makes far fewer people obviously bad. I think that’s the Great Flattening.
Not that AI makes everything identical – something we’re still pretty good at spotting (or maybe some people really do “hope this email finds you well”). And it’s not that creativity disappears. But it dramatically narrows the gap between poor work and perfectly acceptable work.
The floor rises much faster than the ceiling
That’s a strange kind of progress. The average presentation becomes better, the average report becomes clearer, the average image becomes more polished and the average email becomes more coherent. “Quite good” stops being something anyone notices.
Do you even know who in your team writes the best first draft any more? I’m not sure I do. That’s not because everyone suddenly became equally good writers. It’s because AI has compressed what used to be a very visible difference. It’s like smartphone cameras: twenty years ago, taking a genuinely good photograph required expensive equipment and a decent amount of technical knowledge. Today, almost everyone carries a camera capable of producing technically excellent images. Exposure, focus and colour balance have become largely automatic.
That hasn’t made everyone Annie Leibovitz, but it has shifted where the value lies. The scarce thing is no longer technical competence. It’s seeing something worth photographing in the first place.
Excellence might be harder to spot in ‘knowledge work’ than art, but it matters
For years we’ve rewarded polish because polish was rare and expensive. We promoted people who could produce immaculate reports, beautiful presentations and carefully crafted communications because that craft made a visible difference. Increasingly, those things are just table stakes.*
I think the people you’ll value most won’t be the fastest drafters or the slickest presenters; they’ll be the people who improve everyone else’s thinking. The colleague who asks the awkward question that changes the direction of the meeting, or the one who spots the hidden assumption in a beautifully argued report. The person who points out that the recommendation is technically correct but politically impossible. Or who connects ideas from different disciplines in a way that creates something genuinely new. That’s what you’ll notice.
AI can polish an answer, but it still struggles to redefine the question.
Perhaps that’s why that AI-generated sequence stayed with me. It wasn’t because I was worried about actors or directors or the death of creative industries; those are important conversations, but also not really the risk here. Everyone was properly employed and paid; the AI simply gave them a bit more bang for their buck.
Instead, I kept thinking about how quickly something extraordinary had become ordinary. And I knew enough to explore that idea, not just generate a lazy article about how AI is killing the film industry. Not long ago, an AI-generated action sequence would have been The Story. This week, the story (for me) was that I almost didn’t notice it at all. Which probably means I’ve already watched, read and admired plenty of other AI-generated work without realising.
I suspect you have too. That’s why I think this matters; not because AI has suddenly become extraordinary – that’s just an ongoing evolution. It’s because, almost without us noticing, it’s become ordinary. And whenever something ordinary becomes abundant, we start valuing something else instead.
*Table stakes are the minimum amount of money or chips you need just to sit down and play poker. They’re the price of entry. In business, it has come to mean a basic expectation that everyone has, rather than something that gives you a competitive advantage. I could have just said ‘the price of admission’, but I like poker. And tangents.
