Software QA & Testing

Yes. QA can be part of a development engagement with us, a separately scoped service for software that already exists, or ongoing regression coverage. For an existing product, we usually start with a short assessment of risks and current tests, then agree a test plan with you.

We derive tests from your requirements, not from the generated code, because AI-written tests can simply confirm what the code does rather than what it should do. We also look harder at areas generated code can get subtly wrong: authorization, input validation, error handling, duplicated logic and dependencies nobody chose. A code audit is often a useful first step.

It covers the app around them: sign-in, payments, permissions and integrations. The AI behavior itself needs a different kind of testing, with evaluation datasets, answer-quality and grounding checks, tool permissions, failure handling and regression checks when models or prompts change. We scope that as AI Evaluation & Testing, alongside your QA.

Only within the boundary you choose. With Private / Local AI Engineering, models run on infrastructure you control or in an isolated environment we agree with you. With Claude Code / OpenAI Codex Engineering, commercial providers process code under account, data-handling and retention terms agreed before work starts. Where possible, we test with synthetic or masked data rather than live personal data.