Automate BRD to ERD: AI Agent Skills for System Analysts

As a System Analyst (SA), your primary goal is clear: deliver technical document as fast as possible so the development team can start building.

But let's be honest, the manual work is a bottleneck. Transforming a Business Requirement Document (BRD) or Functional Specification Document (FSD) into an Entity Relationship Diagram (ERD), API Specifications, and granular Developer Tasks takes hours, if not days.

Naturally, we look to AI for help. But using AI for this specific workflow comes with its own set of challenges.

The AI Dilemma: Context Loss vs. Over-Engineering

If you've tried simply pasting a massive BRD into an LLM, you already know the pain points:

1.Context Limits: BRDS and FSDs are huge. Feeding them directly into an AI often leads to truncated outputs or hallucinated requirements because the model loses track of the critical details.

2.The "Black Box" Generation: If an AI just spits out the final ERD and API spec in one go, you lose the most critical part of the SA process: human review. As analysts, we don't just need a document; we need to iterate, correct, and align the logic with business realities.

So, what's the alternative? Building a custom RAG (Retrieval Augmented Generation) system with vector databases to process your documents?

No, That is pure over-engineering. You are an SA, not an AI engineer. You don't need a custom pipeline; you need a streamlined workflow.

The Sweet Spot: AI Agent Skilss

This is where Agent Skills come in. Instead of building a system from scratch, I leverage the built-in skill capabilities of modern AI models to guide the AI's behavior precisely how I need it.

I created an agent skill called fsd-analyzer. It is designed specifically for System Analysts to bridge the gap between raw requirements and ready-to-use technical documentation without losing the human-in-the-loop.

Here's how it changes the workflow:

1. Input the Document

You provide the BRD or FSD. Then split them to piece of features or segment with markdown file format. The skill understands the context and structure of standard requirement documents, ensuring no critical details slip through the cracks.

2. Discuss and Iterate (The Human-in-the-Loop)

Instead of asking for the final output immediately, you chat with the AI. You discuss the business logic, question edge cases, and refine the system architecture together. This ensures that the AI isn't just generating text, but acting as an intelligent sounding board that respects your corrections.

3. Generate the Final Output

Once the logic is solidified, you instruct the agent to generate the final deliverables based on predefined templates embedded within the skill. The result?

•A structured ERD

•Detailed API Specifications

•Actionable Developer Tasks ready to be copy-pasted into your Project Management tool (Jira, Monday, Trello, etc.).

No Over-Engineering Required

The best part? You don't need to build infrastructure to use this. Major AI models today including Claude, Gemini, Qwen, Kimi, and Codex already support the concept of agent skills or custom instructions. You just plug the skill in, and you're ready to go.

By using fsd-analyzer, we eliminate the manual typing, avoid the pitfalls of context loss, and maintain complete control over the architectural decisions. It's fast, iterative, and produces exactly what developers need to start coding.

Try It Yourself

If you're a System Analyst, Product Manager, or Tech Lead looking to speed up your technical documentation workflow, I've made this skill open-source.

Check out the repository, load it into your preferred AI model, and let me know how it improves your sprint planning!

🔗 GitHub Repo: https://github.com/rogasper/system-analyst-skill﻿

﻿

Frequently Asked Questions (FAQ)

What is the fsd-analyzer agent skill?
fsd-analyzer is a custom AI agent skill designed for System Analysts. It guides LLMs to read Business Requirement Documents (BRD) or Functional Specification Documents (FSD), discuss the logic with the user, and generate structured technical outputs like ERD, API Specs, and Developer Tasks based on predefined templates.
Can AI generate ERD and API specs automatically?
Yes, but raw generation is risky. Simply asking AI to generate an ERD or API spec form a large document often leads to context loss or inaccurate logic. Using an agent skill like fsd-analyzer allows you to iterate and correct the AI before it generates the final output, ensuring accuracy and maintaining human oversight.
Do I need to build a RAG system to process BRD with AI?
No. Building a custom Retrieval Augmented Generation (RAG) pipeline just to process internal documents is over-engineering for most System Analysts. Modern AI models support agent skills natively, allowing you to achieve structured, context-aware outputs without writing backend code or managing vector databases.
How does AI help in software development documentation?
AI accelerates the translation of business requirements into technical specifications. Instead of manually writing ERDs, API endpoints, and developer tasks. an SA can user AI agent skills to draft these documents in minutes, review them iteratively, and deliver them to the development team much faster.