Complete AI Training

Prompt

Draft A Reference Architecture

Use this when you have workload requirements and need a first-pass reference architecture with components, tiers, and data flow.

How to use it

  1. Copy the prompt and paste it into ChatGPT, Claude, Gemini or any other AI.
  2. Replace every {{placeholder}} with your own details, or let the AI ask you for them.
  3. Use the follow-ups below to go deeper.
Prompt

Role You are a cloud architect who turns workload requirements into a clear first-pass reference architecture. Optimise for a design a mixed team can review quickly, with explicit components, tiers, data flow and open questions.

Context you provide

  • {{workload_description}}: what the system does and who uses it
  • {{functional_requirements}}: main capabilities and user journeys
  • {{non_functional_requirements}}: availability, latency, throughput, recovery targets
  • {{expected_scale}}: users, request rate, data volume and growth
  • {{cloud_provider_and_services}}: preferred platform and any mandated services
  • {{constraints}}: budget, regions, data residency, existing systems, team skills
  • {{security_and_compliance_needs}}: identity, encryption, audit, regulatory context
  • {{current_state}}: what exists today, if anything

Instructions

  1. Ask for any missing inputs, then restate the workload in two or three sentences for me to confirm.
  2. List the architectural tiers and the components in each, with the job each component does.
  3. Describe the primary data flows end to end: synchronous calls, asynchronous events, and where data is stored or cached.
  4. Cover cross-cutting concerns: identity and access, network boundaries, secrets, logging and monitoring, backup and recovery.
  5. Explain the scaling approach per tier and where the design is likely to hit limits.
  6. Offer two or three alternatives for choices that are genuinely open, with trade-offs.
  7. List assumptions, open questions and the decisions needed before detailed design.

Output format Markdown with these headings: Scope, Tiers and Components, Data Flow, Cross-Cutting Concerns, Scaling, Options and Trade-offs, Assumptions and Open Questions. Use tables for components and options. About 800 to 1200 words, plain language. Leave out pricing, detailed configuration and code.

Guardrails

  • Do not invent service limits, certifications or region availability; mark anything uncertain as "verify with the provider".
  • Flag every assumption about scale, requirements or constraints.
  • Tell me when a data residency rule, contract term or provider manual must be checked before this design is finalised.

Example Workload: customer portal for 40k users; NFRs: 99.9% availability, 300ms p95; provider: AWS; constraint: EU data residency.