Skip to main content

ai12z Copilot Platform: Structured Data Integration with Custom Agents

Overview​

The ai12z Copilot Platform is an advanced AI assistant framework designed to streamline complex tasks by leveraging large language models (LLMs) in conjunction with specialized agents. One of its core strengths is the ability to handle structured data efficiently through a reasoning engine that interacts with custom agents. These agents can interface with various data sources such as SQL databases, graph databases, Customer Relationship Management (CRM) systems, Product Information Management (PIM) systems, Content Management Systems (CMS), spreadsheets, Multiple Listing Services (MLS), and a lot more

Table of Contents​

  1. Understanding Structured Data
  2. The Role of Custom Agents
  3. Integration with Data Sources
  4. The Reasoning Engine
  5. Use Cases
  6. Benefits
  7. Getting Started

Understanding Structured Data​

Structured data refers to information that is organized in a predefined manner, often in the form of tables, rows, and columns. This organization facilitates easy storage, querying, and analysis by computers. Common examples include:

  • Databases (SQL, NoSQL)
  • Spreadsheets (Excel, Google Sheets)
  • Data stored in CRM and PIM systems
  • Content in CMS platforms
  • Real estate listings in MLS databases

Structured data is crucial for various industries, enabling efficient inventory management, customer data handling, content organization, property listings, and more.

The Role of Custom Agents​

In the ai12z Copilot Platform, custom agents act as intermediaries between the reasoning engine and external data sources. These agents are designed to:

  • Interpret parameters passed by the LLM reasoning engine.
  • Execute queries against structured data sources.
  • Retrieve and filter data based on user-defined criteria.
  • Return structured responses that the reasoning engine can process.
  • Optionally, bypass the LLM to present data directly to the user interface, such as in a carousel format.

By delegating data-specific tasks to these agents, the platform ensures that the LLM focuses on natural language understanding and generation, while the agents handle data retrieval and manipulation.

Integration with Data Sources​

The platform's custom agents can interface with a variety of structured data sources:

SQL Databases​

Agents can execute SQL queries to retrieve data from relational databases. This is useful for:

  • Fetching records based on specific conditions.
  • Aggregating data for summaries and reports.
  • Updating or inserting records as needed.

Example: Retrieving a list of products that are in stock and within a certain price range.

Graph Databases​

For data best represented as interconnected nodes and edges, agents can interact with graph databases like Neo4j. This allows:

  • Navigating complex relationships.
  • Finding shortest paths or connections between entities.
  • Performing social network analysis.

Example: Identifying mutual connections in a professional network.

CRM Systems​

Agents can access customer data stored in CRM platforms such as Salesforce or HubSpot. They can:

  • Retrieve customer contact information.
  • Update lead statuses.
  • Generate sales reports.

Example: Pulling up a customer's purchase history during a support call.

PIM Systems​

Product Information Management systems store detailed product data. Agents can:

  • Fetch product specifications.
  • Update inventory levels.
  • Manage product categorizations.

Example: Filtering products based on type, price range, and features to help customers find exactly what they need.

CMS Platforms​

Content Management Systems like WordPress or Drupal store articles, blog posts, and multimedia content. Agents can:

  • Search for content matching certain keywords.
  • Retrieve metadata about content pieces.
  • Publish or update content programmatically.

Example: Listing all blog posts related to a specific topic.

Spreadsheets​

Agents can read from and write to spreadsheets, enabling:

  • Data analysis and visualization.
  • Batch updates of records.
  • Importing/exporting data in CSV or Excel formats.

Example: Updating a sales forecast spreadsheet with the latest figures.

Multiple Listing Services (MLS)​

Agents can interact with MLS databases to access real estate listings. They can:

  • Retrieve property listings based on location, price, and features.
  • Update property information.
  • Generate market analysis reports.

Example: Providing a list of available homes in a specific neighborhood within a certain price range.

The Reasoning Engine​

The LLM reasoning engine serves as the brain of the ai12z Copilot Platform. It is responsible for:

  • Understanding user queries in natural language.
  • Determining when to invoke a custom agent.
  • Passing appropriate parameters to the agents.
  • Processing the agents' responses to generate meaningful outputs.
  • Optionally directing the output to bypass the LLM for direct presentation to the user interface.

By collaborating with custom agents, the reasoning engine can handle complex tasks that involve both unstructured language understanding and structured data manipulation.

Use Cases​

E-Commerce Product Filtering​

A customer wants to find "smartphones under $500 with at least 128GB storage and a high-quality camera." The reasoning engine interprets this request and uses a custom agent to query the PIM system, retrieving products that match the criteria.

Process:

  1. User Query: "Show me smartphones under $500 with at least 128GB storage and a great camera."
  2. Reasoning Engine: Understands the criteria and invokes the PIM agent.
  3. PIM Agent: Filters products based on price, storage capacity, and camera features.
  4. Results: A list of matching smartphones is presented to the user.

Real Estate Listings​

A prospective homebuyer is looking for "3-bedroom houses with a backyard in Springfield under $300,000." An agent interacts with the MLS database to find listings that fit these requirements.

Process:

  1. User Query: "Find 3-bedroom houses with a backyard in Springfield under $300,000."
  2. Reasoning Engine: Parses the request and calls the MLS agent.
  3. MLS Agent: Queries the MLS database with the specified filters.
  4. Results: The user receives a list of available properties matching the criteria.

Interactive Product Comparison​

A shopper is interested in comparing laptops and wants to see options in a carousel format with the ability to select items for detailed comparison.

Process:

  1. User Query: "Show me laptops suitable for gaming under $1,500."
  2. Reasoning Engine: Calls the PIM agent to retrieve relevant products.
  3. PIM Agent: Fetches laptops matching the criteria.
  4. Presentation: Results bypass the LLM and are displayed in a carousel with checkboxes next to each product.
  5. User Interaction: The user selects the laptops they're interested in and clicks a checkbox.
  6. Follow-up Query: The system asks, "What is important to you in a laptop?"
  7. User Input: "I need a high-performance graphics card and at least 16GB RAM."
  8. Reasoning Engine: Queries each selected product for full details and generates a comparison.
  9. Results: A detailed comparison table is presented, highlighting how each laptop meets the user's requirements.

Healthcare Data Retrieval​

A medical professional asks for "all patients over 50 with a history of hypertension and no record of diabetes." An agent queries the electronic health records database, and the reasoning engine presents the data in an understandable format.

Supply Chain Optimization​

A logistics manager inquires about "shipments delayed due to customs in the last month originating from Asia." The agent accesses the supply chain management system to provide the relevant data.

Additional Examples​

  • Hospitality Industry: A traveler asks for "pet-friendly hotels with free Wi-Fi and breakfast in downtown Chicago under $200 per night." An agent queries the booking database to find matching accommodations.

  • Educational Institutions: A student looks for "advanced machine learning courses offered next semester that fit my schedule." An agent interacts with the university's course catalog to provide options.

  • Financial Services: An investor wants "stocks in the tech sector with a dividend yield over 2%." An agent accesses financial databases to retrieve a list of such stocks.

Benefits​

  • Efficiency: Quickly retrieves and filters large datasets based on specific criteria.
  • Accuracy: Reduces errors by using precise queries against structured data.
  • Scalability: Handles complex queries across vast data sources without performance degradation.
  • Flexibility: Easily integrates with various data systems and can be customized to specific organizational needs.
  • Enhanced User Experience: Provides users with relevant information rapidly, improving decision-making processes.
  • Interactive Features: Offers advanced UI elements like carousels and comparison tables that enhance user engagement.

Getting Started​

To leverage structured data handling in the ai12z Copilot Platform:

  1. Define Custom Agents: Identify the data sources you want to integrate and develop agents that can interact with them.
  2. Configure the Reasoning Engine: Set up the LLM to recognize when to invoke specific agents based on user queries.
  3. Parameter Mapping: Ensure that the reasoning engine correctly passes parameters to the agents for data retrieval.
  4. Design User Interfaces: Decide when and how to bypass the LLM for direct data presentation, such as using carousels with interactive elements.
  5. Testing: Run simulations to verify that the agents retrieve and return data as expected, and that the user interface behaves correctly.
  6. Deployment: Roll out the integrated system to end-users, providing training as necessary.

By harnessing the power of custom agents and the reasoning engine, the ai12z Copilot Platform offers a robust solution for organizations looking to optimize their use of structured data. Whether it's improving customer interactions, streamlining internal processes, or enhancing data accessibility, the platform is equipped to meet a wide array of needs.