arthai-marketplace

/solution-architect

Design AI solutions for consulting engagements.

Synopsis

/solution-architect <client-name> [--initiative name]

When to use it

Quickstart

/solution-architect acme-corp

What you’ll see: solution designs for the client’s top 3 initiatives by score — technology selection matrices, build-vs-buy analyses, Mermaid architecture diagrams, a Gantt-charted implementation plan, and a 3-year cost model — written to the client’s architecture directory.

Examples

/solution-architect acme-corp                                  # architect the top 3 initiatives by score
/solution-architect acme-corp --initiative "Support Chatbot"   # architect one named initiative

Arguments & flags

Flag Values Default What it does
--initiative initiative name top 3 by score Filter the design work to a single initiative

What it does

  1. Loads context — synthesizes prior outputs from the client’s discovery/ and assessment/ directories: discovery/market-research.md, discovery/competitive-landscape.md, discovery/industry-trends.md, assessment/current-state.md, assessment/ai-readiness.md, assessment/stakeholder-map.md, and assessment/opportunity-matrix.md (the ranking source for “top 3 by score”); extracts objectives, data requirements, integration points, constraints, and budget per selected initiative.
  2. Technology selection matrix — evaluates options per component (LLM provider, vector database, orchestration, data pipeline, monitoring, frontend) on capability, cost, hosting, data privacy level, integration effort, and lock-in risk. Data privacy levels: 1 = public data only (cloud OK), 2 = internal data (SOC 2 required), 3 = PII/sensitive data (encryption at rest/transit required), 4 = regulated data — HIPAA/GDPR/CCPA (on-prem or private cloud required). Level 4 data rules out cloud-only and SaaS options outright, so it’s worth confirming a client’s privacy level before expecting cloud recommendations.
  3. Build vs buy analysis — for each major component: build custom vs SaaS vs open source, compared on cost, time to deploy, customizability, maintenance burden, data control, vendor risk, and skill required, with a recommendation tied to the client’s constraints.
  4. Architecture design — Mermaid system architecture and data pipeline diagrams (chosen from five reference templates: RAG pipeline, ML prediction service, document processing, chatbot/agent, analytics dashboard), an integration points table, and a security and privacy design covering auth, data protection, compliance, and AI-specific risks like prompt injection.
  5. Implementation plan — a Mermaid Gantt build plan, resource requirements table, a risk register with mitigations and owners, and a testing strategy including AI-specific tests (golden-dataset benchmarks, hallucination detection, prompt robustness).
  6. Cost model — 3-year TCO by year, break-even analysis, and a cost scaling projection from pilot to full scale.

A quality checklist runs before output: diagrams render, selections justified, timeline achievable, risk mitigations actionable, security matches the client’s compliance needs.

Output & artifacts

Written to clients/<client-name>/architecture/:

Troubleshooting

Problem Fix
Designs reference data the prerequisites should provide This skill reads discovery and assessment outputs. Run /client-discovery first (writes discovery/), then /opportunity-map (writes assessment/, including opportunity-matrix.md) — both must complete before /solution-architect has what it needs
--initiative doesn’t match anything Use the exact initiative name from assessment/opportunity-matrix.md
A Mermaid diagram doesn’t render Re-ask for the diagram — valid, renderable Mermaid is part of the skill’s quality checklist
Recommendation conflicts with the client’s compliance needs Check the data privacy level assigned (1-4) — regulated data (level 4) forces on-prem or private cloud options