AI-assisted work fails in ways that are easy to miss. Polished output built on incomplete context. Plans built on hallucinated statistics. Strategies built on recommendations no one argued against. The AI is not wrong — the process around it is missing. PDCA (Plan-Do-Check-Act), formalised by Shewhart and Deming, structured human work for the bandwidth era. EC5 is its deliberate evolution for the era in which AI, not human capacity, is the producer of analysis. EC5 — Engineered Clarity 5 — preserves PDCA's loop discipline and extends it with five structural elements PDCA did not need: five steps instead of four (Collect, Check, Consolidate, Challenge, Commit); a defined AI role at each step (Researcher, Critic, Organiser, Sparring Partner, Executor); a nested cycle architecture operating at three levels of abstraction (strategic, execution, operating); two boundary disciplines (the Reality Gate, which protects the method from contamination by unverified real-world material; the Accountability Principle, which protects the method from autonomy drift in multi-agent AI systems); and a documented account of how cycles compound into institutional capability — the Snowball Effect. EC5 is offered as a public methodology in the spirit of PDCA: documented, openly licensed, intended for adoption, adaptation, and critique. The paper specifies the vocabulary, the five steps with their gates and failure modes, the cycle relationships and tree structure, the boundary principles, and the conditions under which EC5 is the wrong tool. A worked example threads through all five steps. Skeletal starting forms for each step's output document are provided in an appendix. The method's claim is structural: clarity in the AI era is not luck, talent, or instinct — it is the product of deliberate structure, built cycle by cycle, by humans using AI with discipline.
Alexandro Wibowo (Tue,) studied this question.