An instrument used in discovery with construction and specialty-trades estimators. The expert tells the story of a recent bid, and the system returns the decisions inside it, structured, ranked and tagged for what to encode and what to keep human.
Ask a senior estimator to list their judgment calls and you get generic answers. Ask about the job they almost walked away from, and every real decision is inside that story: unstructured, and about to be lost in someone's notes.
A story is not a spec. If the consultant reconstructs the logic later from memory, the AI ends up automating a generic idea of the work instead of the expert's actual decisions.
A sample tool that runs anywhere, a live tool for typed stories, and a transcript tool for a full recorded discovery call (recorded with consent). All three use the same server-side extraction, so the API key is never exposed.
Each decision comes back with its trigger, its frame, the real criteria behind it, the move made, and what it leaves behind to learn from, plus a rank and a tag: encode (logic a system could apply) or human (judgment that is relational, political or contextual).
The tags across a whole call become the first draft of an engagement roadmap: what to build, what to leave alone, and in what order.
Four stages between a narrative and a plan an engagement can be scoped against. Click a stage.
A short bid story and the decisions the instrument returns. A sample, not a live call.
Discovery output becomes a build plan, not a pile of notes.
Real decisions live in narratives. Capture the story, then extract the structure.
The rank turns a list of decisions into a sequence.
A process that marks everything as automatable is a sales document, not a diagnosis.
It reports what it cannot do rather than inventing what it can.
The same pattern fits estimating, underwriting, claims triage and procurement. Bring a war story to the first conversation.
📅 Book a call