Architecture Synthesis: Minimum Sufficient Step and ADR Set

The intake brief now defines the decision, KPIs, and risk tolerance the capstone assistant must satisfy. This lesson translates those constraints into architecture decisions. A capstone package is evaluated on decision quality and evidence sufficiency, not on how complete the diagrams appear. Start with the selected decision ladder step for the assistant workload as the first reviewable claim, and justify why it is the minimum sufficient option under the intake constraints. That claim must address two competing risks: insufficient scope can leave critical requirements unmet, while excessive complexity can introduce avoidable failure modes, control overhead, and cost. A defensible outcome may still be to not build, defer, or reshape the solution when the required evidence cannot be produced within the stated risk tolerance or budget constraints.

The working artifact for this lesson is the capstone package update. The three-option paper summary carries over from the intake. This lesson converts the final selection into an architecture decision record set, complete with an index and the rejected alternatives with explicit reopen triggers.

Viable candidates under the same constraints

Start with a set of viable candidates that sit on at least two adjacent steps and plausibly satisfy the same stakeholder KPIs and non-functional requirements. A typical three-option set for an enterprise assistant includes a retrieval-augmented approach that answers from governed corpora, a tool-using approach that calls enterprise systems for actions and live data, and an ...