Here I am, at my first ever React Advanced Conference. It's early in the morning and hundreds of software engineers travelled today to the The Brewery, an event venue in the heart of the city of London. I'm adjusting my just-received lanyard as I slip into the last row in the large hall - just in time. The MCs are stepping onto the stage. I usually prefer sitting closer to the front, but today I arrived a little too late - the room is full, with almost no empty seats left.
The talks cover far more than React itself — from Node.js threads to micro frontends. But among all of them, one topic kept resurfacing, not as a technical curiosity but as an existential question for engineers: the adoption of AI.
In 2025, conversations about AI among engineers are no longer around best code practices and design patterns. They’re tied to job security, visas, promotions, and long-term career planning — especially in a competitive market like the UK. Among my peers, I’ve heard everything from blunt statements like “we won’t need junior developers anymore” to dark jokes about retraining as plumbers, thanks to the advice of the Godfather of AI, Geoffrey Hinton.
Discussion panel: Effective Adoption of AI Coding Roles for Engineers
What tool to use
The panel opened with a question about tooling. The hosts mentioned Copilot, ChatGPT, Gemini, and several others. The consensus is rather clear: as of 2025, the AI landscape is evolving so quickly that locking yourself into a single tool isn’t realistic. Doing so risks missing out on advantages introduced elsewhere almost overnight. Flexibility, rather than tool loyalty, is a strategy.
Does AI increase our productivity?
One experiment shared by the panel stood out.
Two groups were assigned the same feature. One group was allowed to use AI tools; the other wasn’t. Interestingly, the group without AI finished the feature faster. Meanwhile, the group using AI felt more productive.
What struck me wasn’t the result itself, but the disconnect between perceived productivity and actual outcomes — something I see increasingly in real-world engineering teams.
This difference became clearer once the panel broke engineering work into two distinct phases:
- Planning
- Implementation
Senior engineers tend to be more productive with AI, while junior engineers often become less productive.
The reason lies in how these two phases are handled. As you gain experience, more of your effort shifts toward planning — understanding the problem, defining constraints, and choosing an approach — and less toward pure implementation.
The higher-level the task is, and the fewer constraints you give AI, the more directions it can take. If a prompt is vague and only states what you want, AI can easily go down unhelpful tangents.
Senior engineers are typically able to handle the planning phase themselves and then give AI precise, constrained instructions for implementation. Their prompts are detailed, often including examples, clear boundaries, and even commands the AI can run to validate its own output.
Embracing AI in interviewing
The discussion then moved to hiring, where companies are increasingly splitting into two camps.
One camp insists: “You can’t use AI tools. We avoid take-home tasks and prefer in-house interviews to prevent cheating.”
The other camp takes a different view: “Let candidates use AI, and add a stage where we observe how they interact with it — which prompts they write, how they guide the tool, and how they validate the output.”
After all, this is what day-to-day work already looks like for many engineers. In a market like the UK, where interviews are often opaque and high-stakes, this divide says a lot about how companies actually understand modern engineering work. My sense is that organisations in the second camp will move hundreds of steps ahead.
Early adoption of AI
Some companies are going even further, actively aiming to increase the acceptance rate of AI-generated code. Rather than treating AI as a risk to be minimised, they see it as a capability to be developed — provided it’s guided by strong engineering judgment.
Conclusion
For me, attending this conference wasn’t just about React or AI. It was another lesson in how professional life in the UK is evolving — quietly shifting the skills that matter, and rewarding those who can think clearly within uncertain systems.
The takeaway isn’t that AI will replace engineers, but that it changes what seniority means. Value is moving away from writing syntax and toward designing constraints, shaping intent, and validating outcomes.
For junior engineers, the danger lies in using AI as a crutch rather than a tutor. Generating code without understanding the why often leads to hallucinations and dead ends. But for those who master the planning phase, AI becomes a powerful multiplier.
The goal shouldn’t be to write code faster, but to retain the critical thinking needed to steer the machine in the right direction — using these tools not just to speed up, but to scale our capacity to solve genuinely hard problems.