Capstone Intake: Decision, KPIs, and Risk Tolerance
Explore how to develop a clear intake brief capturing decisions, KPIs, and risk tolerance that anchors architecture design and review. Understand how to define measurable success criteria, governance constraints, and oversight requirements for AI systems to ensure compliance and operational rigor. Learn strategies to turn stakeholder inputs into explicit architecture constraints and evidence plans, enabling defensible design recommendations.
The portfolio package from the last chapter gave a review board a rubric for evaluating a proposal. This chapter applies the same review discipline to the capstone design. It develops one capstone assistant from intake through architecture review. A capstone design is difficult to evaluate when the supported decision, KPIs, and risk tolerance are not explicit. Comparing two plausible architectures becomes difficult when success criteria and oversight requirements are undefined. One option may optimize for latency and cost through greater automation. Another may improve auditability through additional review or approval steps. Without a shared measurement plan, there is no consistent basis for selecting between them. A decision to proceed under those conditions reflects preference rather than evidence-based engineering criteria.
Open the intake brief by naming the decision and the measurable outcome that defines success. Frame the decision as proceed, reshape, buy, defer, or stop. A recommendation not to build remains a valid outcome when the proposal cannot meet measurable success criteria or acceptable safety constraints under the organization’s stated limits. That boundary is part of the evaluation criteria. It acts as a quality gate that prevents downstream architecture work from relying on untestable assumptions.
Scope the system as a bounded workflow rather than an open-ended platform. ...