Two kinds of evidence, kept apart. Work delivered for clients, described without naming them. And the tools Clear Build built for its own practice. Where a figure appears, the page says whether it was reported or estimated.
Delivered for clients under engagement. Each is anonymised: the sector is given, the client is not, and no consent to name has been sought.
Clear Build designed, built and verified a five-link proposal chain for a UK professional engineering consultancy, together with the controls that stop it inventing a number. The interesting part is not the elapsed time. It is what stops the machine producing a plausible figure.
Clear Build built a chain of skills that reads a client onboarding questionnaire, matches its questions against the firm's own source documents, and composes and renders a checked response. Where the firm's register cannot support an answer, the chain escalates to a person rather than guessing.
Clear Build wrote a seven-document AI governance suite for a UK professional engineering consultancy whose AI tools it had built, and issued it to the client on 19 August 2026 as a request for comment. It is reference-aligned to ISO/IEC 42001, with ISO 9001, ISO/IEC 27001 and the ICO and NCSC AI guidance in support. It stands on what it is. No time figure is claimed.
Read the case studyTools Clear Build built and runs in its own practice. Not client work.
A set of skills and processes supporting the end-to-end AI audit, from structuring the evidence to producing the finished document. None of it was commissioned by a client and none of it is client work. The point of it is not only speed. It is that AI-produced documents pass deterministic checks before a person sees them, and the check is code, not a second opinion from the same model.
cb-time is a small web application Clear Build wrote for itself. Narrative notes go in, structured billable entries come out, and the invoice is raised into the accounting package without anyone retyping it. None of it was commissioned by a client and none of it is client work.
Not a case study. A walkthrough of the tasks Mike wishes he had had Claude for as an engineer and project manager, and how he would use AI today, knowing what he has learned since.
