Solution Design & Architecture

Translating business problems into Claude-based solutions: choosing between workflow, agentic and augmented-LLM patterns, designing multi-agent orchestration, and tying the design to business value and SLAs.

1. 1.1Translating Business Problems into Claude SolutionsTurn an ambiguous business request into a scoped, feasible Claude design by testing each capability against what language models do well and badly.
2. 1.2End-to-End Architectures and Feedback LoopsDesign the whole path from input to output and back again, choose a reference architecture that fits the workload, and build in the reliability and feedback a demo never needs.
3. 1.3Selecting an Architectural PatternChoose between an augmented LLM call, a workflow and an agent by asking which constraint is tightest, and default to the simplest pattern that meets it.
4. 1.4Multi-Agent Systems and OrchestrationKnow when an orchestrator with subagents is worth its cost, and design the decomposition, result format, conflict handling and failure recovery deliberately.
5. 1.5Decomposing Complex ProblemsBreak a complex process into pieces owned by Claude, by existing systems and by people, then split Claude's share along the lines that make it testable.
6. 1.6Aligning Solutions to Business ValueTie a design to the value the business actually cares about, express it in the business's own units, and build the cost and SLA case on measured numbers rather than averages and hopes.