How AI pair programming is changing my workflow
I've been using AI coding assistants daily for over a year. Here's an honest account of what's changed, what hasn't, and where the real productivity gains are.
Over the past year, the team and I have started using Claude Code properly in the terminal, not just accepting the odd autocomplete suggestion, but actively working with it as part of our development process. Since then the tooling has matured significantly, and the way we write code has genuinely changed. Not in the way the hype pieces suggest, but in ways that are worth talking about honestly.
What actually got faster
The biggest time saving is in the boring bits. Writing test cases, generating TypeScript interfaces from API responses, scaffolding boilerplate for new components, these tasks used to eat 20 or 30 minutes of your day, every day. Now they take seconds. That doesn't sound dramatic, but it compounds. Over a week, I'm probably reclaiming three or four hours of productive time per developer.
The other area where AI genuinely helps is working with unfamiliar codebases. When we pick up a legacy project, a WordPress theme with ten years of accumulated decisions, or a Magento 2 module written by three different agencies, having an AI that can quickly explain what a particular function does and how it connects to the rest of the system is like having a very patient colleague who's already read everything.
What hasn't changed
Architecture decisions. System design. Understanding what a client actually needs versus what they've asked for. These are still entirely human skills, and they're where the real value of development work sits. The AI can write a React component, but it can't tell you whether you should be building a React app in the first place.
Debugging is mixed. AI is good at spotting obvious errors and suggesting fixes for common patterns. But for the genuinely tricky bugs, race conditions, caching issues, problems that only manifest under specific circumstances, you still need a developer who understands the system deeply enough to form hypotheses and test them.
The skill shift
What I've noticed is that the skill emphasis has moved. Writing syntax quickly was never really the bottleneck for experienced developers anyway. The bottleneck was always thinking: understanding requirements, designing data models, choosing the right abstraction. AI hasn't touched that. What it has done is raise the floor. There's a risk here too. Juniors using AI tools can produce code that looks right but isn't, and they don't have the experience to tell the difference. AI isn't a shortcut past learning the fundamentals.
Our actual setup
I use Claude Code in the terminal for everything from inline suggestions to longer planning conversations, planning approaches, reviewing code, working through complex problems. I don't use AI-generated code without reviewing it. Every line still gets read by a human before it goes into a commit. That's not paranoia; it's just good practice. AI makes confident mistakes, and some of them are subtle.
The honest summary
AI pair programming has made us maybe 15-20% faster on implementation work, and it's made some tedious tasks almost enjoyable. It has genuinely changed team sizes. A senior developer with AI can do what used to take three or four people. That's not theoretical, I've seen it on our own projects. What it hasn't done is lower the bar for who can lead that work. If someone's telling you they can replace your senior developers with juniors using AI, they haven't shipped anything complex with it yet. The experience to know what's right, what's wrong, and what questions to ask, that's not something AI provides.
Curious about how this applies to your projects? Get in touch, I'm always happy to talk through the practical side of modern development workflows.
Chris Ryan
Managing Director
17+ years in full-stack web development, most of it leading teams agency-side across e-commerce, CMS platforms, and bespoke applications. Specialises in infrastructure, system integration, and data privacy, with hands-on experience as a Data Protection Officer. Founded Innatus Digital in 2020 to offer the kind of honest, technically-led partnership that he felt was missing from the agency world.