There's a version of the AI story that says developers are on the way out. From where we're sitting, that's not what's happening. What's actually happening is more interesting, and a bit more demanding: the job is changing shape.
We used AI sparingly up until the end of last year. It helped a little, but it got a lot wrong. Then we moved over to Claude Code earlier this year, and that genuinely flipped things on their head. Most of the time now, we're not writing code by hand. We're describing what we want, checking what comes back, and refining it. It's a strange thing to get used to, but it has changed the way we work.
Like having a senior developer sitting next to you
The best way I can describe it is that it's like pair programming with a really experienced developer, someone sitting alongside you to think a problem through with. In the past, if I was stuck, I'd call one of the other senior devs and we'd spend half an hour debugging or planning an approach together. Now I can do a lot of that with the AI instead.
For something technical, like building a connector or a Shopify app, it can read all the documentation, hold all the context at once, and help you write out a plan. It goes further than that now. Using Shopify’s own AI toolkit, we can connect these tools straight into the platform through MCP (Model Context Protocol), a standard that lets the AI work with live, accurate platform detail rather than a general guess at how Shopify behaves. You sense check that plan, and most of the time it's at least along the right lines. You tweak as you go, but the uncertainty that usually comes at the start of a project is largely gone.
When a new brand comes on board, we carry out a health check of their site. AI lets us go even deeper with that review, reading across the whole codebase and surfacing things that are easy to miss. So a brand gets a true picture of what they’re actually sitting on before any work starts, rather than finding the awkward surprises halfway through a project.
You need to know what good looks like first
But a review is only ever as good as the people reading it. Our deep technical expertise and decades of experience is what lets us prompt it properly and, just as importantly, spot when the output is rubbish that won't actually be any use to anyone.
Because it does still get things wrong. Only this morning it started leading me down the garden path. I questioned it, got the familiar "you're absolutely right," and then it went off in another direction entirely. If you didn't already know what good looked like, you'd have no way of catching that. So the idea that you can hand this to someone with no experience and get a reliable result is, for now at least, a fair way off.
Removing guesswork early on
The other shift is how early we can see whether something will actually work. We've got a brand we're working with moving from Magento to Shopify right now with some genuinely complex needs. In the past, working out how to handle those would have meant hours buried in documentation, trying to picture how the pieces fit and hoping we'd got it right. Now we can spin up a working version with AI, connected to the real platform, and see exactly what it does. So we're making calls based on something real in front of us, rather than a best guess this early on. For a brand weighing up a big platform move, that means seeing the difficult parts working, or not working, before committing to the full build.
What this means for developers
What AI is really accelerating is the work that junior and mid-level developers used to do. At Unified, our team is mostly senior and lead developers, so for us it has meant our senior people spend more of their time on the thinking that shapes a project, and less on the mechanical parts. It’s let us grow what we take on by getting more out of the senior expertise we already have in-house.
But it does change what being a developer means. It's less and less about writing the code, because the code will largely get written for you. It's about understanding what you're actually trying to build: what it needs to do, what the user journey is, what "done" looks like. Get that thinking right and the building follows. Get it wrong and no amount of clever tooling will save you.
If you want to talk through what this means for a project you've got in mind, we're always happy to. Get in touch.
I lead our development team on wide range of projects - from Magento 2 storefronts using Hyvä, to bespoke Shopify builds and fully custom headless applications. I’m passionate about combining clean code with thoughtful design to create seamless user experiences and scalable digital products.
