How IG accelerated application modernisation and upgraded 145 applications using AI-native engineering

Rather than assigning engineers to manually write test suites and build pipelines for each repository, Equal Experts built a reusable, automated uplift platform. The team had to collaborate and agree on the ways of working with multiple teams owning these applications.

The team retained only the core seven-point definition of done (DoD) from the initial assessments and engineered three central architectural components:

• A centralised control plane repository: A single operational environment housing shared principles, project registries, and tool configurations, allowing AI agents to work consistently across dozens of services.
• A live program tracker: An automated, filterable tracker reading directly from GitLab pipelines, merge requests, and Jira to reflect actual verified state rather than self-reported progress.
• A custom orchestrator plug-in: Tooling that evaluated target repositories against the DoD, identified gaps, and deployed specialised sub-agents using Claude Code (running Opus for exploratory problem-solving and Sonnet for execution).

The team authored reusable skills encoding engineering philosophy rather than fragile framework-specific rules. For acceptance testing, the engine standardised on Testcontainers to ensure deterministic execution across varying storage and messaging setups. For performance testing, it deployed Gatling modules informed by Honeycomb observability data, targeting the 20% of endpoints handling the vast majority of production traffic. Separate quality-gate evaluation agents verified every change before code reached client review.

Within five months, 145 applications achieved target maturity and were successfully deployed to live production environments.

The engagement began as a targeted effort to upgrade 74 applications but rapidly gained momentum as the team leveraged AI to tackle complex legacy systems. The team built a program tracker to enable transparency and visibility of progress throughout the engagement. The model established a predictable, auditable mechanism for software modernisation at scale. Impressed by early results, IG expanded the scope first to 104 applications, and ultimately added another 41 services. The harness developed and lessons learned during this transformation serve as the foundation for an estate-wide dark factory designed to modernise over 1,000 services.

The increased application maturity enables direct release via CI/CD pipelines, bypassing formal Change Control to accelerate defect resolution and feature delivery. Pipeline and testing upgrades enabled a shift to 24×7 maintenance windows, allowing daytime releases instead of late-night deployments. This improved release maturity, accelerated delivery speeds, and strengthened testing rigour across application teams.

As one of the engineering managers at IG commented: “I wanted to share a final note that we likely wouldn’t have had the time to do a lot of this work without the help of Equal Experts, and we’re in a better place now as a result.”

Anthropic reviewed the AI harness and custom skills developed during the engagement, offering positive feedback for the solution. They noted: “This is one of the more mature implementations we have seen across teams.”