How PwC UK built an AI-first engineering organisation

Introduction

Why we built a world-class engineering team

As clients increasingly expect AI-first delivery, PwC UK's engineering organisation needed a way to scale software development responsibly across hundreds of technologists, multiple delivery teams, and the governance demands of a regulated environment. Our centralised engineering capability, made that possible with Microsoft GitHub Copilot as the platform we chose to power it.

The starting point was a foundational need: a consistent way to build, productionise, scale, and operate technical assets across our engineering teams. That meant a unified approach to the software development lifecycle with security, responsible AI, and governance embedded from the outset, and with Microsoft GitHub Copilot integrated at the core.

We also had an internal challenge to overcome: multiple engineering teams, multiple delivery models, and fragmented ways of working. Scaling AI-first engineering across that footprint consistently, accountably, and at speed, required more than a tool. It required a new operating model, built around the right partnership.

That's why we formed a centralised engineering capability bringing together nearly 1,000 technologists, software engineers, data scientists, and AI specialists to help translate emerging innovations into practical business outcomes.

The timing was deliberate. The team was designed to be AI-first from day one not to retrofit AI onto legacy ways of working. It sits within a broader network of engineers across PwC and powers our Forward Deployed Engineering (FDE) business, enabling teams to prototype at velocity alongside clients.

Why Microsoft GitHub Copilot: meeting enterprise demands with enterprise-grade tools

Enterprises don't just need AI tools. They need enterprise-grade tools and the ability to manage them as the landscape evolves. That means being able to manage a broad and fast-changing model garden, update to the latest models without rebuilding the surrounding controls every time and support a range of users from professional developers to business users building lightweight tools all within consistent observability and guardrails.

After evaluating multiple solutions, we chose Microsoft GitHub Copilot. It offered the best ecosystem for our needs integrating across our development environments and control frameworks, while providing continuous access to new models as they come out.

Most importantly, Microsoft GitHub Copilot gave us what we call "Safe Speed", the ability to develop at scale across a broad model garden, supported by PwC's governance and oversight framework. Speed without sacrificing governance. Innovation without compromising accountability. For a regulated business and auditor, that combination wasn't a nice-to-have. It was essential.

The solution

Engineering the shift: tools, standards and a new way of working

Adopting AI coding assistants was the starting point, not the strategy. The tools alone don't transform an engineering organisation. What does is the operating model, standards, and culture you build around them. We approached the challenge through three pillars: Capability, Control, and Culture, underpinned by observability and risk management. Together, they define how we've made AI-first engineering work at scale — responsibly, consistently, and accountably.

• Capability — Equipping our people to lead the shift
  AI-first engineering is as much a human shift as a technical one. Our priority from day one was making sure our engineers could use the tools confidently, accountably, and at the standard our clients expect. We invested in structured training and certification, not only on the technology itself, but on how to apply it responsibly. We established a consistent set of engineering principles and ways of working designed for an AI-first environment. And we started to redefine how software gets built, moving from traditional sprint cycles to a continuous flow of build, test, and deploy that matches the pace AI now enables.
• Control — Engineering speed, responsibly
  As a regulated business and an auditor, we have to move both quickly and responsibly. That constraint shaped every decision we made. We built Microsoft GitHub Copilot into a unified control plane: when a new agent, model, or release becomes available, we have the protocols and guardrails in place to ensure it is adopted appropriately and accountably. These controls extend across the full software development lifecycle, not just the coding environment, giving us confidence that speed never comes at the expense of governance. Consolidating our codebases into a single GitHub organisation was a critical early step. It improved access, reduced duplication, made governance enforceable at scale, and reduced costs, turning a fragmented landscape into a foundation we could build on with confidence.
• Culture — Learning by doing, and sharing what we learn
  We didn't get every decision right first time and that's exactly why this experience is valuable. Managing fragmented platforms early on slowed us down before we consolidated. Adapting to a continuous build-test-deploy model required engineers to unlearn habits built over years. Building confidence that AI was there to support, not replace, our people required clear intent and communication. Establishing trust in AI-generated code required deliberate effort, not assumption. Each of these lessons is now embedded in how we operate and in how we guide our clients through the same journey.
• Observability and risk management — Keeping “Safe Speed” honest
  Underpinning all three pillars is a discipline of observability and risk management designed for the AI-first era. By capturing baseline metrics from the outset, we track adoption, usage, and impact in real time across the engineering organisation. We apply consistent guardrails as new models and agents enter the garden, gauge return on investment with precision, and maintain auditability end-to-end supporting the governance standards expected of a regulated business.

The results

Turning AI-first adoption into tangible benefits

The shift to AI-first engineering is no longer theoretical for us. It's measurable in speed, quality, and the way our engineers spend their time. We can now deliver full-stack engineering up to 30% faster, with more value delivered sooner across both our internal platforms and our clients. Our engineers are now able to focus on enhancements, value adds, and new features, with Microsoft GitHub Copilot now generating the majority of the boilerplate code in our platforms, which frees them to drive the future innovations.

Our Forward Deployed Engineering (FDE) business can now prototype with clients at a faster pace than before, helping to turn AI-first capability into tangible client outcomes on the ground.

Crucially, this speed has not come at the expense of quality or control. The standards and guardrails we embedded during adoption mean we can integrate new agents, models, and releases confidently and apply governance across the full software development lifecycle, not just the coding environment.

But the impact goes deeper than productivity metrics. We no longer hand-code everything. Microsoft GitHub Copilot now handles the bulk of coding development and first-reviews, freeing our engineers from repetitive work and transforming our innovation cycle. The most valuable shift: our engineers now spend more time alongside clients, designing better business outcomes, not buried in code. Embedding AI-first coding standards has enhanced both agility and control.

And finally, by capturing baseline metrics, we've been able to track progress and gauge return on investment with greater precision and insight.

The opportunity

We've walked the journey. Now we're ready to walk with you.

That's our advantage — and it can also be yours.

From technology selection and architectural decisions to governance frameworks, to the cultural and behavioural shifts that determine whether AI adoption succeeds or stalls, we've been through it. We know what works, what doesn't work, and why. Whether you're at the start of your AI-first journey or already in motion, we can help you move faster, avoid the missteps we encountered, and realise the full benefit of enterprise-grade, AI-augmented engineering. Talk to us about your AI-first engineering journey.