The Driver and Vehicle Licensing Agency (DVLA) explored AI coding assistants to improve software delivery while supporting experienced engineers. Working with CGI, DVLA tested a team-level approach, delivering a new service ahead of schedule and enabling developers to take on more complex work with confidence.
Client’s challenge: Balancing innovation and craft
The DVLA maintains driver and vehicle records in the UK and delivers digital services at national scale.
After two years of using AI coding assistants, DVLA faced a clear challenge. The tools offered speed and capability but introduced a period of change. Many engineers had spent years refining their craft and took pride in writing code. At the same time, DVLA needed to improve efficiency and build more digital services.
Early use of AI was cautious and inconsistent. Some engineers were sceptical, questioning its impact on quality and roles. Others used it for small, low-risk tasks. DVLA chose not to mandate usage, focusing instead on maintaining system integrity while supporting its people. The challenge became how to adopt AI coding assistants in a way that respected engineering expertise while enabling meaningful progress.
Benefits and outcomes: Scaled adoption and delivery
The approach marked a shift from individual use to coordinated, team-level adoption of AI coding assistants:
- Service delivered ahead of schedule – completed two weeks earlier than the estimated 3 months (17% efficiency)
- Live production service – sending around 30,000 reminders daily
- Increased confidence – engineers able to take on more complex work and unfamiliar technologies
- Strong adoption – majority of surveyed users report daily use and positive experience
- Ongoing learning – internal community supports knowledge sharing and continuous improvement
DVLA continues to build on these insights, refining how AI coding assistants are used to support developers, while balancing their use with developer judgement to maintain quality, and adapt to new ways of working.
The solution: Team-level AI experiment
To move beyond individual experimentation, DVLA collaborated with CGI to explore how AI coding assistants could be applied consistently at team level.
Together, they tested a structured approach within a live, greenfield project to deliver a driver reminder service. The service processed daily data and sent notifications via email, SMS or post based on customer preferences. This provided a practical environment to apply AI tools across the full delivery lifecycle, moving beyond earlier use in small-scale, individual tasks or isolated experimentation.
The team introduced product requirement prompts to translate backlog tickets into detailed, structured instructions for AI tools. These prompts defined not only what needed to be built, but how to implement it, helping to reduce variation in how developers used the tools. Shared engineering standards and repository-level instructions were also applied to guide more aligned outputs across the team.
All engineers followed the same process, creating a controlled environment to test how AI could support delivery at scale. Alongside this, DVLA continued to build an active internal community, sharing experiences, techniques and evolving best practice. This combination of structure and collaboration helped teams engage with the tools in a more consistent and informed way across the delivery process.
Final outcome
The engagement showed how DVLA could move from individual experimentation with AI coding assistants to a more consistent, team-level approach. By testing the tools within a live service, DVLA built greater confidence in how AI could support software delivery while maintaining developer judgement and quality.
The experience has also given engineers new ways to work, learn and take on more complex tasks. DVLA is now using these insights to continue refining its approach to AI coding assistants and support wider adoption across its engineering teams.
Working with CGI allowed us to explore how AI coding assistants could be applied at a team level. We had the right level of trust between our teams, which allowed us to experiment, learn quickly and develop a structured approach together. The collaboration helped us successfully deliver a live production service ahead of schedule, while also giving our engineers new ways to innovate, share ideas and take on more complex work with confidence.
Head of Software, DVLA
Learn more
To find out more about how we can help you harness the transformative power of AI:
