When you think of ASOS, the London-based fashion retailer, clothing probably comes to mind first. But every part of that customer experience, from browsing to checkout, depends on an engineering team that ships code and releases new features to the ASOS app and website every day across more than 150 markets and 17 million customers.

But with 700 engineers working across 3,000+ repositories in Azure DevOps, that pace was hard to sustain. Engineers regularly worked in codebases they didn’t own, often in languages they didn’t use daily, slowing how quickly teams were able to deliver value to their customers.

ASOS needed a development environment that made its codebase easier to navigate and gave engineers the tools to move quickly once they found the right place to work. The team, led by Distinguished Engineer Dylan Morley and Principal Software Engineers Chris Newark and Lewis Holmes, saw an opportunity to bring their repositories and AI-powered development workflows into a single platform.

The idea for us is to empower our developers to go faster. We need to get more features out to our customers and GitHub Copilot is enabling us to do that at speed. - Dylan Morley, Distinguished Engineer

Moving 3,000 repositories to GitHub

ASOS moved all of its repositories from Azure DevOps to GitHub’s AI-powered platform, unifying their codebase in a single AI-powered environment where Copilot has context across every repo. "With over 3,000 repositories, Copilot helps us do a number of things. It helps us navigate them. It helps us understand the concepts of them and enables engineers to jump between repositories and really get up to speed as quickly as possible," says Morley.

The impact on day-to-day work was immediate. Engineers can now easily get up to speed with previously unfamiliar services, which used to require days of onboarding or conversations with teams that once felt siloed. Instead of spending their time understanding unfamiliar code, ASOS developers now spend more time problem-solving and driving innovative ecommerce experiences, like Virtual Try-On.

With the migration in place, ASOS focused on getting Copilot into the hands of every engineer. Newark used Copilot usage metrics to identify power users early, then turned them into champions who could show peers what was possible in their own codebases. That peer-led rollout helped ASOS reach over 90% Copilot adoption across the engineering team within months.

GitHub Copilot brings back the fun. It allows engineers to focus on problem solving instead of getting bogged down with technical restrictions. - Chris Newark, Principal Software Engineer

Embracing multiple models

ASOS gives engineers the flexibility to choose the models that work best for the task at hand. And as new models become available through the platform, they can switch between them without shifting workflows — whether they're working in an IDE or running tasks from the command line with Copilot CLI.

That flexibility matters in an industry where the tooling changes week to week. When a new model or capability becomes available, ASOS engineers can pick it up through the same platform they're already using.

We can pivot quickly with Copilot. This platform gives us the variety we need to keep up with this insane pace of change within the industry. - Chris Newark, Principal Software Engineer

Prototyping at a new pace

With GitHub Copilot, ASOS engineers can now delegate more complex development tasks to AI agents, helping them prototype at a pace that wasn’t possible before. They can also take on work that would have required skills the team didn't have.

For example, when a team of back-end engineers needed to build an internal stock reconciliation tool, they estimated two months of work. "This team was all back-end focused with no front-end capability," says Morley. "With Copilot cloud agent, we delivered in nine days what we'd estimated at two months." The team even rebuilt the tool three times in that window, tweaking the prompts and regenerating the system to refine the output.

These productivity gains weren’t exclusive to engineering teams . Holmes trained product managers to use Copilot, enabling them to generate working prototypes directly. What used to take weeks of back-and-forth between product and engineering now happens in minutes.

With agents in the loop, engineers can experiment more freely, taking on ideas that would have been deferred or skipped in the past.

There's always 100 things that we want to build. With GitHub Copilot, we get to do more of them. - Chris Newark, Principal Software Engineer

Solving the code review bottleneck

With 90% of the company’s engineers using GitHub Copilot, ASOS was producing more code than ever — not just because engineers were faster, but because agents were generating code too. That meant more pull requests and a growing review bottleneck. PR first approval times peaked at 56 hours, so the ASOS team enabled Copilot Code Review across every repository to take the first pass on code reviews.

“Copilot spots issues we missed. We’re confident it improves both velocity and code quality,” says Holmes. The team also built a custom agent that guides developers through the review process, which saves them time and keeps them focused.

Now standard across ASOS repositories, Copilot Code Review provides fast feedback on pull requests and has brought PR first approval time down.

Spending less time on PRs with Copilot Code Review means engineers can focus on the work that matters and ship products as fast as possible. - Dylan Morley, Distinguished Engineer

Orchestrating agents at scale

Even with faster coding and reviews, ASOS engineers were still losing time to repetitive tasks like bug fixes and documentation updates, limiting their opportunities to apply deeper skills.

The team built a custom agent infrastructure with Mission Control: Bug Basher, available inside every IDE. Engineers delegate well-defined tasks (like squashing bugs) to agents running parallel in the cloud, and then they can review the output when it suits them. They don’t even need to work synchronously: Newark often assigns a bug to the agent, goes to a meeting, and then assesses the output when it’s convenient. Embedding agents directly into the SDLC made bug resolution faster and more reliable without sacrificing quality.

Holmes also wired Copilot into fleet management tooling to cut down on manual updates, enabling engineers to write custom prompts and roll them out across thousands of repositories at once, which reduces context switching and keeps engineers in the flow.

With Mission Control and custom agents, one team builds an agent and 700 engineers benefit. Bugs, documentation, test coverage: all delegated, all running in parallel while our engineers focus on what matters. - Dylan Morley, Distinguished Engineer

Delivering more features to customers

Agentic workflows and the ability to choose the best model for the job have fundamentally changed how ASOS engineers spend their time. Tasks that once took an hour can now often be completed in 10 minutes of agent runtime.

These changes have unlocked a new level of productivity: faster implementation, quicker prototyping, and more creative problem solving. Engineers now spends less time writing code and more time designing solutions. The result: more products in front of their customers at speed, with ASOS delivering trends before anyone else.

Fashion is about identifying trends and being quick to the market. GitHub and the agents are enabling us to do that at speed. - Dylan Morley, Distinguished Engineer