System Design Requirements: A Prompting Workflow for Your AI Assistant

Many software engineers fail the system design interview before they draw a single box. The most common mistake isn't a poor architecture choice; it's diving into a solution without first clarifying the problem's requirements, scope, and constraints. Practicing this crucial first step is difficult because it requires a dialogue partner who can be deliberately ambiguous. An AI interview assistant is the perfect tool for this, but only if you prompt it correctly.

This guide provides a repeatable workflow for using an AI assistant to simulate the requirements clarification phase. You'll learn how to set up your AI as an interviewer who reveals information only when asked, forcing you to develop the muscle memory for asking the right questions about users, scale, and trade-offs before committing to a design.

Why Standard AI Prompts Fail for System Design Practice

If you simply ask an AI to "give me a system design problem," it will likely provide both the question and a complete, detailed solution. This is counterproductive. Reviewing a perfect answer doesn't build the skill of navigating ambiguity under pressure. As noted in system design preparation guides, interviewers evaluate your reasoning and ability to handle unclear initial specs. The goal of practice isn't to memorize solutions, but to master the process of discovery.

The Practice Contract: Prompting Your AI to Act as an Interviewer

To start a useful practice session, you need to give the AI a specific role and a set of rules. This 'practice contract' ensures the AI acts as a sparring partner, not a cheat sheet. You can use this prompt in a tool like Acedly's Assessment Assistant or Live Copilot to begin a focused session.

“

Act as a system design interviewer for a Senior Software Engineer interview. The session is 45 minutes. Start by giving me one ambiguous design prompt, such as 'Design a URL shortener' or 'Design a news feed'. During the simulation: 1. Only answer the specific questions I ask about requirements. 2. Do not volunteer information, suggest technologies, or correct my design. 3. If I make an assumption about scale or features, ask 'What led you to that assumption?' 4. Keep a private list of functional and non-functional requirements I successfully clarified. When I say 'END SIMULATION,' provide feedback on my requirements gathering. Score my clarification process from 1-5 and list three critical questions I missed.

This prompt establishes clear boundaries. It forces you to lead the conversation, just as you would in a real interview. You can adjust the role (e.g., "Mid-level Software Engineer") and time constraints to match the interviews you're targeting.

A Step-by-Step Workflow for Clarification Practice

Once you've set up your AI interviewer, follow a structured approach to uncover the system's hidden requirements. For this example, let's assume the AI gave you the prompt: "Design a URL shortening service."

1. **Restate and Define Scope:** Start by confirming the core goal. Ask: "So, the goal is to take a long URL and generate a shorter, unique alias that redirects to the original URL. Is that correct? Are there other key features I should
2. **Clarify Functional Requirements:** Ask about specific user actions. For the URL shortener, these are crucial: Can users create custom aliases? Do links need to expire? Do we need to support link editing or deletion? Must users have an
3. **Estimate Scale (Non-Functional):** Ask questions that inform your back-of-the-envelope calculations. "What is the expected number of new links created per month? What is the anticipated read-to-write ratio (i.e., redirects vs
4. **Define Performance and Availability (Non-Functional):** Uncover latency and reliability targets. "What is the maximum acceptable latency for a redirect? Is this a global service? What is the availability target, for example, 99.9%?"
5. **Probe for Constraints and Edge Cases:** Ask about what's explicitly out of scope. "Should the system handle analytics, like click tracking? Do we need to consider security aspects like preventing malicious links?"

Reviewing Feedback and Iterating

After you've gathered what you believe are sufficient requirements, end the simulation. The AI, following your initial prompt, will now provide feedback on your questioning process. Pay close attention to the "missed questions." Did you forget to ask about regional data residency? Did you fail to clarify the consistency model for analytics?

The goal is to run this drill repeatedly with different prompts (e.g., "Design a ticketing system," "Design a leaderboard"). Each time, you'll internalize the framework, making the process of clarifying requirements second nature. This structured practice transforms the most ambiguous part of the system design interview into a predictable, manageable sequence.