Architecture Prompts: AI-Assisted System Design and Engineering

Overview

Architecture decisions are some of the most consequential work engineers do. A poor data model, a flawed distributed systems design, or an ill-considered microservices decomposition can take years to pay down. AI assistants can accelerate architecture work — not by replacing architectural judgment, but by rapidly generating options, stress-testing assumptions, and producing detailed technical specifications for human review.

This section contains four specialized prompt libraries covering the major domains of software architecture work:

Section Contents

System Design Prompts

25+ prompts for designing complete systems from scratch.

Covers: distributed systems, database design, API design, event-driven architectures, and system design interview preparation. Use these when you need to go from requirements to architecture quickly, or when you want to explore multiple design options before committing.

Go to System Design Prompts →

Scaling Prompts

25+ prompts for designing scalability into systems and handling growth.

Covers: horizontal scaling strategies, database sharding, caching tiers, CDN architectures, queue-based load leveling, and auto-scaling configurations. Use these when your system is approaching limits or when you need to design for 10x-100x current scale from the start.

Go to Scaling Prompts →

Migration Prompts

25+ prompts for planning and executing large-scale technical migrations.

Covers: database migrations, monolith-to-microservices, cloud migrations, framework upgrades, and API versioning strategies. Use these when facing a major migration project where the risk of mistakes is high and the need for a structured plan is critical.

Go to Migration Prompts →

Cost Optimization Prompts

25+ prompts for reducing infrastructure and operational costs.

Covers: cloud spend analysis, right-sizing, reserved vs. spot capacity, database query optimization, CDN cost reduction, and FinOps practices. Use these when cloud bills are unexpectedly high or when you need to build a cost-conscious architecture from the start.

Go to Cost Optimization Prompts →

How to Use Architecture Prompts Effectively

Providing the Right Context

Architecture prompts work best when you provide:

1. Scale requirements: Current traffic, projected growth, peak vs. average load
2. Team constraints: Team size, expertise, existing technology stack
3. Non-functional requirements: Latency SLOs, availability targets, compliance requirements
4. Cost constraints: Monthly budget, key cost drivers
5. Existing architecture: What you currently have, what's working, what's not

The Architecture Prompt Pattern

Most effective architecture prompts follow this pattern:

Context:
- What you're building / what you have
- Current scale / projected scale
- Key constraints (team size, cost, deadline, compliance)

Ask:
- Design [specific system/component]
- Compare [option A] vs [option B]
- Identify risks in [this design]
- Create migration plan for [this change]

Output format:
- Architecture diagram description (for mermaid)
- Component list with responsibilities
- Data flow
- Risk/trade-off analysis
- Implementation sequence

Validating AI Architecture Output

AI-generated architecture should always be validated:

1. Stress-test assumptions: Ask the AI to critique its own proposal
2. Check for known anti-patterns: Distributed transactions, chatty microservices, N+1 query patterns
3. Estimate costs: Use AWS/GCP pricing calculators for proposed infrastructure
4. Review with your team: Especially engineers who have implemented similar systems
5. Prototype before committing: For novel architecture choices, prototype to validate before full implementation

Architecture Decision Records

When using these prompts to make architecture decisions, generate an ADR:

Generate an Architecture Decision Record (ADR) for:

Decision: [what was decided]
Context: [the situation requiring a decision]
Options considered: [list from prompt exploration]
Decision: [chosen option and rationale]
Consequences: [good and bad consequences of this decision]

Format: MADR (Markdown Architectural Decision Records) format

Prompt Quality Principles

The prompts in this section are designed around these principles:

Specificity over generality: "Design a rate limiting system for an API with 10,000 req/s" produces better output than "design rate limiting."

Trade-off focused: The best architecture prompts ask for trade-offs, not just recommendations. There is rarely one right answer.

Iteration friendly: Start broad, then narrow. Use follow-up prompts to drill into specific components.

Reality-grounded: Include your actual constraints — asking for a system with no budget, no team size limits, and no existing technology is an interesting thought experiment but not useful for production work.

Related Sections

• UI Prompts — Component and design system prompts
• Design System Prompts
• Postmortem Framework — For architecture decisions that went wrong
• Chaos Engineering — For validating your architecture's resilience