Essential session storage is required to sign in. Optional analytics remain off unless you allow them.
Choose a narrow work task such as support triage, account-meeting preparation, policy drafting, onboarding planning, document review, or knowledge lookup. Build a product-discovery package and thin vertical slice. Requirements: (1) observe or model the current workflow with a defined user, trigger, inputs, decisions, exceptions, and outcome; (2) write an opportunity statement, non-goals, success metric, and critical failure condition; (3) build one end-to-end path using authorized or synthetic data; (4) show a user-review state with accept, edit, reject, or escalate behavior; (5) document what is simulated, what is real, and what would need hardening before production.
The user, trigger, current pain, inputs, decision, exceptions, opportunity statement, non-goals, and success criteria are specific enough to evaluate with a real stakeholder.
The prototype completes one authentic path from input to reviewable output rather than presenting an isolated model demo or generic chat interface.
The project clearly distinguishes real versus synthetic data, prototype shortcuts, failure behavior, user control, and the next evidence needed before production work.
Use this README structure to make the work reviewable.
DevPath fetches a read-only, size-capped snapshot. Submitted code is never executed.
# An evidence-backed AI workflow prototype ## Goal Choose a narrow work task such as support triage, account-meeting preparation, policy drafting, onboarding planning, document review, or knowledge lookup. Build a product-discovery package and thin vertical slice. Requirements: (1) observe or model the current workflow with a defined user, trigger, inputs, decisions, exceptions, and outcome; (2) write an opportunity statement, non-goals, success metric, and critical failure condition; (3) build one end-to-end path using authorized or synthetic data; (4) show a user-review state with accept, edit, reject, or escalate behavior; (5) document what is simulated, what is real, and what would need hardening before production. ## Acceptance criteria - [ ] Workflow and product evidence: The user, trigger, current pain, inputs, decision, exceptions, opportunity statement, non-goals, and success criteria are specific enough to evaluate with a real stakeholder. - [ ] Thin vertical slice: The prototype completes one authentic path from input to reviewable output rather than presenting an isolated model demo or generic chat interface. - [ ] Honest boundaries: The project clearly distinguishes real versus synthetic data, prototype shortcuts, failure behavior, user control, and the next evidence needed before production work. ## Verification - [ ] Document installation and run commands. - [ ] Record automated checks and their exact commands. - [ ] Add representative output, screenshots, or a short demo where useful. - [ ] Confirm failure paths and known constraints. ## Decision log ### Decision title - Context: - Choice: - Alternatives considered: - Tradeoffs: ## Evidence - Link each rubric criterion to the file, test, or artifact that demonstrates it. ## Known limitations - List what is intentionally out of scope and what should be improved next.