AI Insights

Godel AI engineers introduced Claude Code agents into NDM’s test automation workstreams, more than doubling automated test output to deliver customer change faster. 

NDM needed to accelerate customer change without adding delivery risk. The goal was to increase reliable coverage, shorten release cycles and prove measurable throughput gains inside the teams and metrics already in place. 

NDM’s constraint was not ambition; it was capacity. 

Limited SDET availability made it harder to sustain automated test coverage at the level needed for faster releases. Coverage had held at around 65% across key parts of the application suite, but after crunch periods where engineers were pulled into manual testing, coverage fell to 25–30% in parts of the stack, and the remaining tests became harder to trust. 

Adding headcount alone would not close the gap quickly enough, so NDM turned to Godel’s forward-deployed engineers to rebuild pace and confidence. NDM protected quality by swarming manual testing around each release. But that protection came at a price: slower feature delivery, longer release cycles and a reduced pace of change for the 3 million members across healthcare, education, caring and charity. 

NDM did not need more experimentation. It needed a delivery model that could restore automation capacity fast, prove impact in live workstreams and create the foundation for agentic delivery.

The answer wasn’t headcount. It was a new way to deliver. 

Godel’s AI engineers embedded Claude Code agents directly into NDM’s daily test automation work, creating a reusable delivery system rather than a one-off AI trial: 

• Test Case Creation Skill 

• Web UI auto-test Agent 

• Web API auto-test Agent 

• Auto-test PR Review Agent 

• Failed Tests Fixing Agent 

The result: 133% faster test automation output 

The productivity jump traditional delivery couldn’t unlock. The impact was measured against NDM’s existing sprint data, not a separate innovation benchmark. The gain showed up in the same delivery rhythm the business already used to track progress.

Product	Measure	Before agents	With agents	Uplift
Aurora	Tasks per sprint	3.6 (sprints 48–64)	8.4 (sprints 65–69)	+133.3%
Elysium	Story points per sprint	9.6 (sprints 1–24)	17.8 (sprints 25–28)	+85.4%

On Aurora, Claude Code agents accelerated API and UI auto-test creation and took on pull request review. On Elysium, they stabilised the full auto-test suite and created every viable API auto-test. Both results were measured like-for-like, using the same teams and delivery units. 

In business terms, the shift breaks the old trade-off – protect quality or move faster. Coverage that had fallen to 25–30% in parts of the stack is now measured at 91% of viable UI test cases and 76% of viable API test cases on Aurora, with every viable API auto-test created on Elysium. That puts the regression safety net back into the automated suite, releases manual QA effort that had been absorbing engineering capacity, and removes a major brake on how quickly NDM can put change in front of members. 

“Godel has both breadth and depth across technologies, and the level of service means if something isn’t going well, I know about it quickly, usually before I know it’s a problem myself. I didn’t want a body shop. I wanted a partner who would help take the platform in the direction it needed to go, and who isn’t afraid to tell me I’m wrong and suggest a different way of doing things.” 

Simon Rowe, CTO at NDM 

The rationale is throughput and reliability. Faster, more dependable automation shortened release lead times and moved NDM toward its strategic goal of fully agentic development. Simon Rowe credits Godel with being “first to the table with a more structured plan” for working through the SDLC and the safeguards needed before agents can run through it. 

That momentum is already visible inside NDM, where teams dedicate time each week to building new AI skills and agents across the business. 

For NDM, the outcome is more than faster test creation. It is a working proof that AI agents can be introduced into live engineering delivery with measurable impact, trusted safeguards and a clear route from productivity gain to customer value. 

“This project shows what happens when AI is built directly into delivery, not treated as a side experiment. By bringing Claude Code agents into NDM’s test automation workstream, we helped more than double automated test throughput, creating a faster route to market and a clearer path to return.” Joe Wolski, CTO at Godel Technologies