Lloyds Banking Group: Copilot and GitHub Copilot at scale

The bank wanted generative AI to become part of everyday work for tens of thousands of colleagues, not a tool used by a small early group, and wanted engineers to move faster on large code conversion and development tasks. Doing this inside a regulated bank also meant keeping AI within its existing security, identity and governance controls.¹

Lloyds started with 300 Microsoft 365 Copilot licenses and in March 2025 committed to 40,000. To build habits it trained more than 10,000 people, set up a network of about 1,000 volunteer "flight instructors" and ran weekly promptathons to share good practice. Nearly 5,000 UK engineers were using GitHub Copilot by October 2025, and more than 10,000 by June 2026.

In June 2026 the bank signed a multi-year agreement to deploy Microsoft 365 E7 across the Group. Microsoft describes E7 as combining Microsoft 365 E5, Microsoft 365 Copilot and Agent 365, with Microsoft Entra Suite and Defender, Intune and Purview capabilities. Lloyds plans one self-service agent for all colleagues to find systems, information and answers, plus further agents for specific colleague and customer journeys, all controlled through Agent 365.¹²

In our view, the reusable lesson is how Lloyds built usage: volunteer champions, weekly prompt practice and broad training took it from 300 licenses to near-universal use among license holders. Banks with large legacy codebases can also aim GitHub Copilot at code conversion first, where time savings are easier to show. The figures need context. The 46 minutes comes from a survey of 1,000 users, The Register called it strangely specific and noted the bank has not said how the saved time is used, and the usage rates have no published definition. The code results are single projects. The E7 bundle mainly makes sense for an organization already standardized on Microsoft security and identity; a smaller bank could start with Copilot Chat and a few paid licenses.