A software consultancy (BBD) shares lessons from an accelerated delivery project with a major financial institution, where AI was used across the entire SDLC to compress eight months of work into three weeks. Key findings: AI accelerates creation but not enterprise integration; prompt engineering is a specialist skill; verification is essential as outputs are often incomplete rather than wrong; and the last 20% of production-readiness still takes 80% of the effort. The biggest enabler wasn't AI itself but cross-functional JAM sessions combining all stakeholders. The conclusion is that engineering discipline, senior oversight, and human judgment remain critical — AI replaces slow delivery, not engineers.
Table of contents
Eight months of work delivered in three weeksThe magic and mess of using AI across the SDLCThe last 20% is still 80% of the workThe delivery model that actually worksAI-native, enterprise-literate engineering teamsAI won’t replace engineers. It will replace slow deliverySort: