Gemini

Gemini sits in our internal workflow rather than in anything we hand over. We use it mainly for test data — generating realistic, varied datasets that respect the shape and constraints of a client’s domain, which is otherwise slow manual work and tends to produce data too uniform to catch edge cases. It is also useful for compressing long defect threads into something a stakeholder can act on, and for drafting the first version of test documentation that an engineer then corrects. Where a client already runs their test matrices and reporting in Google Sheets and Docs, the Workspace integration means we can work inside their existing artefacts instead of exporting into ours. We do not let it near production data, and nothing it generates reaches a client deliverable without an engineer reviewing it — a plausible-sounding test case that asserts the wrong thing is worse than no test at all.