Essential session storage is required to sign in. Optional analytics remain off unless you allow them.
Deliver an end-to-end AI product case study and working critical path suitable for a controlled production pilot. Requirements: (1) discovery evidence, target job, success metric, non-goals, and positioning statement; (2) architecture covering identity, tenant or role boundaries, data minimization, source freshness, model routes, tools, authorization, audit logs, and fallback; (3) product-quality and security threat model with accountable owners; (4) representative evaluation suite and release gates; (5) latency, reliability, and cost budgets with routing, caching, and failure behavior; (6) analytics from request through user decision to downstream outcome, with privacy-safe instrumentation; (7) pilot cohort, support, communication, incident runbook, feature pause or rollback plan, and a post-pilot decision; (8) if tools write data, include explicit approval, validation, idempotency, auditability, and reversal design.
The discovery, workflow, product promise, technical architecture, controls, and launch plan tell one coherent story about a valuable user outcome and its constraints.
Evaluation, privacy, security, governance, authorization, tool boundaries, reliability, cost controls, observability, incident response, and rollback are practical and proportionate to risk.
Analytics, feedback, pilot results or decision criteria, user communication, source and model versioning, and a next-step decision show how the product would improve safely after launch.
Use this README structure to make the work reviewable.
DevPath fetches a read-only, size-capped snapshot. Submitted code is never executed.
# A production AI product launch package ## Goal Deliver an end-to-end AI product case study and working critical path suitable for a controlled production pilot. Requirements: (1) discovery evidence, target job, success metric, non-goals, and positioning statement; (2) architecture covering identity, tenant or role boundaries, data minimization, source freshness, model routes, tools, authorization, audit logs, and fallback; (3) product-quality and security threat model with accountable owners; (4) representative evaluation suite and release gates; (5) latency, reliability, and cost budgets with routing, caching, and failure behavior; (6) analytics from request through user decision to downstream outcome, with privacy-safe instrumentation; (7) pilot cohort, support, communication, incident runbook, feature pause or rollback plan, and a post-pilot decision; (8) if tools write data, include explicit approval, validation, idempotency, auditability, and reversal design. ## Acceptance criteria - [ ] End-to-end product judgment: The discovery, workflow, product promise, technical architecture, controls, and launch plan tell one coherent story about a valuable user outcome and its constraints. - [ ] Production quality and trust: Evaluation, privacy, security, governance, authorization, tool boundaries, reliability, cost controls, observability, incident response, and rollback are practical and proportionate to risk. - [ ] Evidence-driven operating model: Analytics, feedback, pilot results or decision criteria, user communication, source and model versioning, and a next-step decision show how the product would improve safely after launch. ## 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.