If something looks wrong, you don't scramble for a hotfix. You flip the switch back instantly. The risky moment just disappears.
Nikhil Thakare
Head of DevOps, Cloud, and QA @ The AA
Case Study
Customer Snapshot
The AA is one of the UK’s most recognized brands and is often called the country’s fourth emergency service. It serves millions of members across breakdown recovery, insurance, loans, savings, accident assist, and its driving school. More than 95% of its sales now happen digitally, and a large share of breakdown calls are reported online rather than over the phone. When that much of the business runs through a screen, every change to the software is a change customers feel.
The AA partnered with GlobalLogic to build Cruise Control, its internal FeatureOps platform, powered by Unleash. Here is what changed.
Results
- Zero major incidents caused by a release since adopting FeatureOps
- 35% more deployments every month, year on year, with operations getting calmer instead of noisier
- 20% lift in conversion on key customer journeys, driven by data-backed rollouts and experimentation
- A per-feature audit trail of who released what, when, and to whom, replacing the generic release change record
- Legacy CMS migration tested safely in production, with no customer-facing content issues
The challenge: deploy and release were the same event
Like most organizations, the AA had welded two things together that did not need to be welded: shipping code and exposing it to customers. Every deployment was a release. That meant every release carried the full risk of a deployment, so the team leaned on code freezes, sign-offs, change advisory boards, and fixed release windows to feel safe.
It worked, but it was slow, and it could be frightening.
“The coupling between deploy and release as one inseparable event is the thing that made changes at the AA slower and sometimes quite scary.” — Nikhil Thakare, Head of DevSecOps, Cloud and QA, The AA
The AA operates in a highly regulated environment with contractual and regulatory commitments that do not move. It needed to ship faster and stay in control of exactly what went live and when. Bureaucracy could not deliver both. So the team asked a different question: what if deploying and releasing did not have to be the same thing at all?
Why Unleash
Two reasons stood out when the AA evaluated feature flag platforms.
The first was extensibility. The team wanted to wrap Unleash in a custom layer and follow one consistent pattern across the business, managed centrally rather than reinvented by every team.
The second was privacy, which mattered to the data and privacy stakeholders from day one.
“The data never leaves our servers, and evaluation happens server-side for operations involving PII. This helped us get buy-in from our stakeholders, mainly in the data and privacy team.” — Nikhil Thakare, The AA
Local, in-application evaluation meant sensitive customer data stayed inside the AA’s infrastructure while the team still got real-time control over behavior in production.
The solution: FeatureOps, made native to the AA
Cruise Control is FeatureOps made native to the AA. Unleash is the engine; Cruise Control is the whole experience the teams interact with. GlobalLogic built it as a paved road so the right way to ship is also the easiest way to ship.
“Code can make its way all the way to production switched off, sitting in a dormant state, invisible. Deploy no longer equals releasing features to customers.” — Nikhil Thakare, The AA
With the link broken, product owners decide who sees a feature and when, not the deployment pipeline. Teams roll out gradually: internal users first, then a slice of external traffic, then everyone once the data supports it. And when something looks wrong, there is no scramble.
“If something looks wrong, you don’t scramble for a hotfix. You flip the switch back instantly. The risky moment just disappears.” — Nikhil Thakare, The AA
GlobalLogic delivered this through a platform engineering approach, built in three layers:
- Developer portal. Engineers manage flags from Autopilot, the AA’s Backstage instance, alongside their pipelines and security scans. A Backstage plugin for Unleash did not exist, so the team built one and open-sourced it. Production toggles still flow through the standard Unleash release process, keeping controls intact.
- FeatureOps engine. Unleash sits here, with a custom SDK wrapper the AA’s own product teams built. It adds Angular-native dependency injection, easier flag testing, and automatic lifecycle management, so flags do not become permanent clutter.
- Digital platform. The infrastructure, CI/CD, governance, and secrets management underneath. Teams self-onboard to Cruise Control through a GitHub PR backed by a Terraform module, following GitOps practice. Tokens are vended and injected automatically from AWS Secrets Manager. New services come up feature-flagged by default, “the AA way,” from day one.
For the operations team, the payoff is operational, not just developmental.
“For the operations team, it’s a genuine kill switch, which enables fewer major incidents and a clean audit trail of every change made into production.” — Nikhil Thakare, The AA
The outcome
The numbers tell the story the team set out to write: faster delivery and more control at the same time.
“We’re deploying 35% more each month, year on year, and things are getting calmer, not noisier. We’ve had zero major incidents caused by a release. That’s the number I’m most proud of.” — Nikhil Thakare, The AA
Cruise Control also gave product teams something they had wanted for a long time: the ability to roll out on their own terms, watch real production data, and act on it. That experimentation muscle translated directly into a 20% lift in conversion on key journeys, and let the AA migrate a legacy CMS while testing in production without exposing customers to content issues.
Adoption grew the way good platforms do, by being useful.
“Once teams see the benefits, they use it more and more. The more teams use it, the more they want to use it.” — Lorne Stalker, Engineering Director UK & Ireland, GlobalLogic
What’s next
The AA is extending Cruise Control beyond its digital components into core IT systems, including COTS and self-built platforms, so company-wide change programs can be governed end to end across the whole stack. The team is also building Cruise Control into its GenAI software delivery lifecycle, giving AI coding tools the skills to use feature flags correctly and making FeatureOps a first-class part of how AI-assisted code ships.
“Adopting technology for technology’s sake never works. There has to be a business reason behind it, and then there has to be a ‘so what.'” — Lorne Stalker, GlobalLogic
For the AA, the “so what” is clear: ship faster, sleep better, and keep millions of members moving.