Almost everyone I know who has built AI into their day-to-day work is busier than they have ever been.
That's worth holding onto, because it doesn't fit neatly with most of the conversation about what AI means for jobs.
Some of the worry is completely understandable. The technology is moving quickly, it can already do things that would have seemed ridiculous a few years ago, and plenty of roles and tasks will change significantly as a result.
I'm just not convinced the story is as simple as AI arriving and people disappearing. What I see looks much more like a reshuffle.
A junior developer with the right AI tools can now attempt work that might previously have needed somebody more experienced. A mid-weight developer can produce the kind of output we once associated with someone senior. And the senior developer shifts sideways, spending less time writing every line and more time on architecture, reviewing output, making judgement calls and directing several streams of work at once.
The same is happening outside development. A junior designer can explore more concepts, produce more variations and work across mediums that might once have needed several different people.
That changes jobs. It changes what we expect from people, and it probably changes the shape of teams. Whether it removes the people is a separate question, and a much less certain one.
AI raises the floor
One of AI's real strengths is breadth. It can code, write, analyse, summarise, create images, work through documents and hold more context at once than any individual could keep in their head.
We see this in our own work all the time. AI can often get you 80 or 90% of the way through something big, very quickly.
In software, that can mean someone getting an entire application working when they would never previously have had the time or the breadth of knowledge to build it. Huge numbers of people are now "vibe coding" apps into existence, and some of them are genuinely impressive. They work, they look polished, and they prove an idea incredibly quickly.
Getting something working is one milestone. Getting it ready for production is another, and the gap between the two is where most of the hard work lives: architecture, security, permissions, resilience, data handling, deployment, testing, monitoring, maintainability, and all the awkward edge cases nobody thought about on day one.
Above all, that last stretch needs judgement.
Someone has to know whether a shortcut is perfectly reasonable for a prototype or a serious problem waiting six months down the line. Someone has to spot the architectural decision that looks fine in isolation but will make the next stage unnecessarily painful. Someone has to notice when the AI has confidently produced something that simply isn't right.
That expertise hasn't gone anywhere just because the early stages got easier. There's a reasonable case that it has become more valuable. Far more people can now get close, which leaves the people who can reliably finish with more to do.
Very good at being good
This is the part I think is easiest to miss when people talk about what AI can do.
AI is very, very good at being good. And it can be good at a scale no person can match.
An average developer producing an average amount of code is one thing. A system that produces reasonably good code across an entire application, in the time it takes a person to get through a fraction of it, is something else altogether.
The same applies creatively. Someone who has never operated a camera, built a 3D scene or edited a commercial can now make something surprisingly convincing, without a production budget. That's brilliant. It opens creative work up to people who never had access to it.
It also comes at a cost. When it becomes this easy to produce something good enough, we end up with an awful lot of mediocrity. LinkedIn posts that all seem to come from the same place. Websites that are perfectly competent and almost impossible to tell apart. Reports that are well structured and say very little. None of it is bad, exactly. That's the problem. There is so much of it that the work with something to say gets harder to find.
Put any of it next to the work of a genuinely exceptional creative and the difference shows. Taste. Restraint. Originality. Knowing what to leave out. Understanding why one choice is right and another is merely fine. Having a point of view.
AI can write a novel. That doesn't make it Stephen King. It can create a striking image. That doesn't make it Van Gogh. I'm deliberately using extreme examples, and the point holds further down the scale too. When competent becomes instant and almost free, competent stops setting anyone apart. The value moves to the things the tools find hardest.
There is no single "AI"
It's also worth remembering that there is no single thing called "AI" that is equally capable at everything.
We use different models for different jobs because they have different strengths. One might be better at working through a complex code problem, another at extracting information, another at visual work, another at handling a very large amount of context. Sometimes we run the same problem through more than one because the first answer isn't quite right.
Even with the best tools available and a good sense of which one to reach for, you don't get perfection. Models misunderstand context. They miss nuance. They make assumptions. They can produce something technically correct that still feels completely wrong.
That creates another kind of expertise: knowing which tool to use, how far to trust it, where to check it, and when to ignore its answer altogether.
The better the tools get, the more tempting it is to assume that judgement is no longer needed. My experience so far has been almost exactly the opposite.
If AI is replacing us, why are we all so busy?
Which brings me back to where I started.
If adopting AI simply left people with less to do, you'd expect the heaviest users to be gradually running out of work. That isn't what I'm seeing. The people leaning hardest into it are exploring more ideas, building more things, taking on work that wouldn't previously have been commercially practical, and finally getting to the problems that sat on the "one day" list because nobody had the time.
A lot of that time comes from the jobs nobody wanted in the first place. Developers have often put off writing documentation. Now AI can draft most of it from the code itself, and a developer checks it over. The time saved goes straight into the work they actually came to do.
The capacity gets filled. It always has, and I don't see AI changing that.
That doesn't prove there will be no job displacement. There will almost certainly be disruption, and roles built around repeatable or fairly standard tasks will come under pressure. That's exactly why I think we need to be careful when the conversation jumps straight from "AI will change employment" to "fewer people will be working". The first is very likely. The second is a much bigger assumption, and treating the two as one makes it harder to prepare well.
For the person doing that work, AI is also a way to move up or across. That's the reshuffle I mean.
Expertise still sets the ceiling
None of this means everybody's job is safe. Some work will disappear. Some roles will change dramatically. We may decide certain tasks simply don't need a person doing them any more.
Even so, "AI replaces humans" misses most of what's interesting. AI is giving many more people access to capabilities that used to need specialist knowledge, significant resources or a lot more time. And when it can take more people 80 or 90% of the way, the people who can recognise and deliver the remaining 10% may become more important.
The developer who understands the architecture. The creative with genuine taste. The accountant who knows what the numbers mean for the business. The adviser who reads the person sitting opposite them. The specialist who can tell when an answer looks right and isn't.
The most useful thing any of us can do is learn the tools well enough to move with them: use them properly, understand where they fall short, and put our own effort into the parts they can't do.
If you're in Guernsey, the AI Sprint is running free sessions until 12 November, including some on the island's workforce and the human skills that matter as AI develops. DEXM are the Sprint's community tech partner, and built its website.
AI raises the floor enormously. Expertise still sets the ceiling.
I expect plenty of jobs to change shape because of that, and I expect people to move upwards, sideways and into roles that don't quite exist yet. From where I'm sitting today, though, I'm far less convinced the people disappear along the way.
By Josh Gosselin, Founder, DEXM