I've been building websites and apps for over fifteen years. In that time I've seen frameworks come and go, jQuery give way to React, and servers turn into clouds. None of it changed the day-to-day work as much as AI has in the last couple of years.
There's a lot of noise about it, from "developers will be gone by next year" to "it's just autocomplete". Neither matches what I see at my desk, so here's my honest take.
Where it really helps
The biggest win is the boring work. Boilerplate, migrations, test scaffolding, converting one data format into another, writing the fiftieth form validation rule: an AI assistant does these in seconds, and I review them in a minute. That's time I get back for the parts of a project that need thought.
It's also a brilliant second pair of eyes. Asking it to review a function, explain an unfamiliar library or suggest why a query is slow often gets me to the answer faster than searching through old forum threads. It doesn't always get it right, but it's usually pointing in a useful direction.
And it has made me braver. I'll try a new language or tool for a small job now, because the cost of getting started is so much lower.
Where it falls down
AI is confident, and confidence isn't the same as being right. It will happily invent a function that doesn't exist, or write code that works for the example and breaks on real data. It doesn't know your business, your customers or the reasons behind a decision made three years ago.
- It can't decide what to build. Understanding what a client actually needs is still a conversation between people.
- It doesn't own the result. If something goes wrong at 2am, someone has to understand the system well enough to fix it.
- It's only as good as the brief. Vague instructions get vague, plausible-looking code.
- Security and privacy need a human who knows where the data goes and who can see it.
How I use it
I treat AI like a very fast junior developer who has read everything and remembers most of it. I give it clear, small tasks, I read everything it writes, and I never ship code I don't understand. The judgement, the architecture and the responsibility stay with me.
AI doesn't replace knowing what good looks like. It makes it more valuable.
What it means for clients
For the people I build for, the upshot is mostly good news. Projects move faster, there's more time for design and testing, and features that used to be out of budget, like a smart search, a content assistant in the CMS or automatic summaries, are now realistic for a small business.
What hasn't changed is the need for someone who understands the whole picture: your goals, your users, and how all the pieces fit together. That's the job, and AI has made it a more interesting one.