ChatGPT Prompts for System Design: Architecture Documents Fast
ChatGPT prompts for system design and architecture documents. High-level designs, component diagrams, trade-off analysis, and technical decision records.
The Prompt
Act as a Staff/Principal engineer who has designed distributed systems handling 1B+ requests per day.
Create a system design document for:
System: {what you're building — describe the feature or system in one paragraph}
Scale requirements: {requests/day, data volume, latency requirements, availability SLA}
Team size and experience: {affects architecture complexity — small teams need simpler systems}
Existing infrastructure: {what's already in place that this system must integrate with}
Key constraints: {budget / compliance / technology choices / timeline}
Design goal: {MVP / production-ready / proof of concept}
Output:
1. Requirements clarification (10 clarifying questions you'd ask in a real system design interview — and answers based on context)
2. Capacity estimation (back-of-envelope math: storage, bandwidth, compute requirements)
3. High-level architecture (components and their interactions — described in text and pseudocode, no diagrams needed)
4. Data flow (step by step: what happens from request to response)
5. Component deep-dives (for the 2-3 most complex components — design decisions and alternatives)
6. Trade-off analysis table:
| Decision | Option A | Option B | Recommendation | Rationale |
7. Failure modes and mitigation (top 5 ways this system can fail and how to handle each)
8. Scaling plan (how to scale from 10K to 1M to 100M users — what changes at each tier)
Constraints:
- Start simple — recommend the least complex solution that meets the requirements
- All capacity estimates must show the math
- Trade-off analysis must present genuine alternatives — not strawmen
- Failure modes must include both technical failures and operational failures
Variables to fill in
-
{system description}What you're building — one detailed paragraph -
{scale requirements}Requests per day, data volume, latency SLA, availability SLA -
{existing infrastructure}What this system must integrate with -
{team and constraints}Team size, experience, budget, compliance, technology choices -
{design goal}MVP, production-ready, or proof of concept
How to use this prompt
- Run this before writing any technical spec — it surfaces requirements you didn't know you had
- Use the capacity estimation section in your engineering proposal to justify infrastructure costs
- Share the trade-off analysis with your team before making architecture decisions
- Use the failure modes as your incident runbook starting point
Start simple — complexity is a tax
Every component you add to a distributed system is a potential failure point, a maintenance burden, and an onboarding challenge for new engineers. The most common system design mistake is over-engineering for scale you don’t have yet. The prompt explicitly recommends the least complex solution that meets requirements — which is the correct starting point for any system.
Capacity estimation grounds the design in reality
System design without numbers is philosophy. ‘100K requests per day’ means 1.2 requests per second at peak — a load that a single $10/month VM can handle comfortably. ‘100M requests per day’ means 1,200 requests per second — which requires a fundamentally different architecture. Back-of-envelope math is the step that separates theoretical design from practical engineering.
Failure modes are where senior engineers earn their salary
Junior engineers design for the happy path. Senior engineers design for failure. The top 5 failure modes section forces you to think about what happens when a database is unreachable, when a queue fills up, when a service deploys a bad version, when traffic spikes 10x in 30 seconds. Designing for these scenarios before they happen is vastly cheaper than recovering from them in production.
Related free tools
- Free AI Prompt Generator — build structured prompts instantly
- AI Token Counter — estimate API costs before long runs
- AI ROI Calculator — measure AI’s impact on developer productivity