LIFE WITH AI: PRACTICAL

AI “Content Engineering” – reform the team with ASAA (AI System Analysis Assistant)

The AI System Analysis Assistant (ASAA) is an AI-powered toolkit designed to perform “Content Engineering” on existing documents for a specific software system. It acts as an intelligent expert system, enabling teams to efficiently exchange information and accelerate understanding of complex systems.

Core Functionality

ASAA processes a wide range of existing documentation to extract structured knowledge, enabling seamless collaboration and information flow among team members across different roles. The system also serves as a learning platform for onboarding new members.

Documents Analyzed:

  • System Requirement Side:
    • Emails, meeting minutes, business requirements, interface agreements.
  • Architectural/Design Side:
    • Architectural requirements, designs, system designs, detailed implementation designs.
  • Testing Side:
    • Test plans, automation scripts, and testing data requirements.
  • Feedback:
    • Customer feedback and other relevant inputs.

Key Benefits by Role

  1. System Requirement Side (BA, PM, Solution Architect):
    • Facilitate quick access to technical details and enable seamless business discussions.
    • Examples:
      • BA asks: “What are the key system constraints mentioned in the interface agreement?”
      • PM queries: “Summarize the customer feedback related to scalability.”
  2. System Developer/Implementation Side (System Engineer, Data Engineer, QA):
    • Simplify understanding of business requirements and technical design details.
    • Examples:
      • Data Engineer asks: “What fields in the database schema align with the revenue-related business logic?”
      • System Engineer queries: “How is error handling designed in the interface agreement?”
  3. QA Engineers:
    • Obtain details for specialized test requirements, test data preparation, and automation support.
    • Examples:
      • QA queries: “What test automation scripts are linked to error handling?”
      • QA asks: “Provide specific test scenarios for validating the system’s scalability features.”
  4. New Team Members:
    • Serve as a training resource to quickly understand the overall system and dive into any specific details.
    • Examples:
      • New hire asks: “What’s the high-level architecture of the system?”
      • Follow-up: “Explain the interaction between modules in the interface agreement.”

Detailed Implementation Plan

1. Data Ingestion and Parsing

  • Objective: Import and analyze all system documents.
  • Implementation Steps:
    1. Use OCR tools (e.g., Tesseract) for scanning non-digital documents.
    2. Employ text parsing tools (e.g., spaCy, Hugging Face Transformers) for document classification and extraction.
    3. Store parsed data in a knowledge graph or structured database for easy querying.

2. Semantic Understanding

  • Objective: Enable ASAA to interpret questions in natural language and map them to the correct data or documents.
  • Implementation Steps:
    1. Train NLP models using relevant datasets (e.g., project documentation, requirements, QA plans).
    2. Use tools like LangChain or OpenAI API to develop semantic search and question-answering capabilities.
    3. Implement document summarization for long documents using LLMs (e.g., GPT models).

3. Expert System Development

  • Objective: Build a system capable of answering specific technical and business queries.
  • Implementation Steps:
    1. Develop a knowledge graph linking various documents, concepts, and relationships.
      • Example: Link “error handling” in requirements with specific design implementations.
    2. Integrate tools like Neo4j or TigerGraph for graph-based querying.
    3. Enable multi-modal input support (e.g., text, voice, images) for ease of use.

4. Collaborative Interface

  • Objective: Provide a unified platform for team members to access, query, and exchange information.
  • Implementation Steps:
    1. Build an intuitive UI/UX with input fields for natural language queries.
    2. Integrate communication tools (e.g., Slack, Microsoft Teams) for seamless discussions.
    3. Add role-specific dashboards for BA, PM, engineers, and QA.

5. Continuous Learning and Updates

  • Objective: Keep the system updated with the latest documents and evolving requirements.
  • Implementation Steps:
    1. Implement a document ingestion pipeline with version tracking.
    2. Use active learning techniques to fine-tune models based on feedback.
    3. Set up automated notifications for changes in requirements or test cases.

Technologies and Tools

  • Data Ingestion: Tesseract, Apache Tika.
  • NLP Processing: Hugging Face Transformers, spaCy.
  • AI Models: OpenAI GPT, Azure Cognitive Services.
  • Knowledge Management: Neo4j, TigerGraph, Elasticsearch.
  • Automation and Integration: LangChain, Zapier, Slack Bots.
  • Collaboration Tools: Microsoft Teams, Confluence.

Example Use Case : Expert for the system of the “Adventure Works Cycles” business

Adventure Works Cycles is a global manufacturing company specializing in the design, production, and sale of bicycles and related accessories. The company operates across various functional areas, including production, sales, distribution, and customer service.

The business operation & database models (including REST APIs), are leveraged to create a realistic and fully documented system environment. This environment serves as input for “Content Engineering,” which facilitates the development of an AI-based helper system called ASAA.

This ASAA-like tool is designed as an expert chatbot tailored for the “System Definer,” a role typically fulfilled by business-side team members. The purpose of ASAA is to provide a comprehensive understanding of how the existing system operates. It assists in identifying the components and data that would be affected if new features are introduced. Additionally, ASAA evaluates whether the current system contains sufficient data to support proposed features or improvements, ensuring informed decision-making during the system enhancement process.

All documents (from overview, requirements, high-level design, detailed design for each components, and the Java code & database DDL, integration test plan, .etc) of the are feed to this ASAA-like expert system for “Adventure Works Cycles” business …

This ASAA-like expert system exposed in Telegram messaging as Chatbot (can also be exposed in messaging system like Whatsup, Microsoft Teams, .etc), then the team member can directly ask system related question just as chat with a well trained system expert, no worry about how to find the original designers / developers of the system who already left the team….

(FYI: if want to try this sample expert system, simply join Telegram, search the virtual person “LifeWithAITest_bot”, then you can chat with it about any details about how the system of “Adventure Works Cycles” is running …)