How We Built an AI-Native SDLC with Claude at JAJA
Kindgeek engineers have been part of JAJA’s tech team for six years. Since June 2026 they have been rebuilding how the whole team plans, builds, tests and ships software around Claude.
Industry
Services
Hire AI Engineers, AI Engineering
Timeline
Since June 2026, ongoing
Team Size
8 people
JAJA Finance is a UK fintech and credit card provider with around 760,000 customers. Kindgeek has partnered with JAJA for six years, building products from the ground up and supporting their development with a consistent team. Our embedded AI engineers now lead Claude adoption on JAJA’s own account. Claude entered the engineering workflow in June 2026, and since August the team has also been building Airi, a Claude-powered customer service assistant. This case study covers the delivery side: an AI-native SDLC with Claude in requirements, coding, code review, testing and QA, used daily by JAJA’s own developers and QA engineers, with every change still checked by people.
Challenges
Faster delivery without a bigger team
JAJA’s roadmap kept growing. Hiring more people is slow, and every new developer needs weeks before they are productive. The quicker route was to change how the existing team works.
Faster coding alone would only move the bottleneck
When developers write code faster, the pressure moves to review, testing and QA. Unless those steps sped up as well, faster coding would only create a longer queue in front of every release.
Claude across the lifecycle
“We have worked with Kindgeek for years, but what the team has brought recently stands out. Building Claude into how we plan, build, test and ship software has changed the way our team delivers.” Wayne Cavanough, CTO/CIO, JAJA
Requirements
Drafts and refines user stories and acceptance criteria from briefs, so stories reach developers clearer and earlier.
Product owner approves every story.
Development
Claude Code helps write, refactor and document code inside JAJA’s repositories.
The developer owns every change.
Code review
Runs a first-pass review of each pull request and flags issues before a human reviewer looks.
A human reviewer approves every merge.
Testing
Generates unit and integration tests together with new code and updates existing tests when code changes.
Developers review the tests; CI runs them.
QA
Turns acceptance criteria into test cases and regression checklists and helps reproduce and describe bugs.
QA engineers run the final checks.
Release
Changes ship through JAJA’s existing CI/CD and approval gates.
What We Built
01
Embedded Engineering Capacity
Experienced Kindgeek engineers working as part of JAJA’s tech team, inside its sprints and its rituals.
02
Claude in Requirements
User stories and acceptance criteria drafted and sharpened with Claude, so work reaches developers clearer and earlier.
03
Claude in Coding and Review
Developers write and refactor code with Claude Code, and every pull request gets a Claude first-pass review before a human reviewer takes over.
04
Claude in Testing
Tests are generated together with the code and kept up to date as the code changes, so coverage grows without slowing delivery.
05
Claude in QA
Test cases and regression checklists come from acceptance criteria, reducing manual preparation, so QA engineers focus on cases requiring human judgment.
06
Enablement and Guardrails
Pairing, shared practices and clear review rules, so JAJA’s own team uses Claude confidently and safely on JAJA’s own Claude account.
Results
2x
Faster sprint delivery
Work moves through the whole pipeline faster, from a clear story to a merged, tested change, because every slow step got shorter, not because anyone works longer.
40%
Less QA effort per story
Tests arrive together with the code and test cases come straight from acceptance criteria, so QA engineers spend less time preparing and more time on what needs human judgment.
40%
Fewer bugs on prod
Every change still goes through human review, automated tests and JAJA’s existing checks, and about 40% fewer bugs now reach production. The speed came from removing waiting and rework, not from skipping steps.
77%
Of devs use Claude daily
JAJA’s developers and QA engineers now use Claude as part of how they work, whether or not we are in the room. The skills stayed with the team. Everything runs on JAJA’s own Claude account. JAJA keeps control of the tooling, the data and the roadmap.