The Corticon AI Assistant leverages artificial intelligence to help you explore, generate, and better understand the assets in your Corticon rule projects. With it, you can ask the AI questions, generate new rules from plain text instructions or code snippets, or make requests about the currently selected rule asset. Potential uses range from simple tasks, such as generating a summary of a rulesheet, to more context-aware inquiries, such as understanding how an attribute is used in a ruleflow or creating new rules based on your business requirements.

The Corticon AI Assistant enhances rule project development and maintenance by providing many benefits, including:

  • Improved understanding of rule assets
  • Automatic generation of project documentation
  • Creation of rules from plain text instructions or code
  • Insight into the impact of planned changes
  • Easier onboarding of new team members

Seamlessly integrated into Corticon Studio, the Corticon AI Assistant brings the power of AI directly into your rule modeling experience.

Prerequisite: You Need Your Own AI Account

To use the Corticon AI Assistant, you must have an active account with an AI service provider -either OpenAI or Azure OpenAI. You will need to obtain an API key from your provider and enter it into Corticon Studio. The Assistant does not work without this, and each organization is responsible for managing its own AI subscription and keys.

Getting Started with the AI Assistant

The Corticon AI Assistant integrates seamlessly into your rule modeling workflow, but it does require a one-time setup before you can begin. These simple steps will help you connect the AI Assistant to your organization’s chosen AI provider and ensure everything is ready for productive use.

Before you begin:

To use the Corticon AI Assistant, you must have an active account with OpenAI or Azure OpenAI, and the required credentials from your provider.

  • OpenAI users: You will need your API key.
  • Azure OpenAI users: You will need your API key and your Azure OpenAI endpoint URL.

Make sure your organization has set up this account and you have access to these details.

To connect and access the AI Assistant in Corticon Studio:

  1. Open Corticon Studio on your computer.

  2. From the top menu, select Windows > Preferences.

  3. In the Preferences panel, go to Progress Corticon > Corticon AI Assistant.

  4. Choose your API:
    • For OpenAI: Enter your API key.

    • For Azure OpenAI: Enter both your API key and the Azure endpoint URL.

  5. Click Apply and Close to save your settings.

To open the Corticon AI Assistant panel in Corticon Studio

  1. Go to the top menu and select Windows > Show View > Corticon AI Assistant.

  2. If Corticon AI Assistant is not shown in the initial list, select Other… instead. In the window that appears, scroll through the options, select Corticon AI Assistant, and then click Open.



Once opened, the AI Assistant panel will appear in your workspace. You can keep this panel open as you work, and close or reopen it at any time as needed. The panel is ready for you to enter prompts, ask questions, generate rules, or troubleshoot your rule assets.

Using the Corticon AI Assistant

  1. Open the Vocabulary, Rulesheet, or Ruleflow of interest in your rule project. The AI Assistant will automatically reflect the currently open rule asset in Corticon Studio:

  2. Enter your question or request about the rule asset:

  3. Press Send to submit your request to the AI.
  4. Review the AI-generated response:

To help you get started with AI, the Corticon AI Assistant offers a set of prompt suggestions tailored to each rule asset type. Clicking a prompt suggestion sends the corresponding question or request directly to the AI:

How the AI Assistant Works

When you click Send to submit a question or request to the AI, the Corticon AI Assistant sends the following to the AI:
  • A serialized form of the currently selected rule asset.
  • A system prompt explaining how to interpret the rule asset.
  • Your prompt, which is the question or request you are asking the AI.

All three elements are included in a single request to the AI, and the AI-generated response is then displayed.

The Corticon AI Assistant is stateless. Unlike ChatGPT, it does not maintain context between prompts sent to the AI. Each question or request you submit is processed independently, without awareness of previous interactions.

Note: The Corticon AI Assistant uses OpenAI GPT-4o-mini. The total size of the prompt, including the rule asset definition, plus the AI-generated response must not exceed the token limits of GPT-4o-mini. This is primarily a concern for ruleflows with 50 or more rulesheets. To use the AI Assistant effectively with very large ruleflows, consider breaking them into smaller subflows and using the AI Assistant on individual subflows.

Disclaimers

  • Policies regarding the handling of data provided to the AI Assistant are between you, the user, and OpenAI. For details, refer to the OpenAI API Platform policies. Progress Software is not responsible for how OpenAI processes data provided through the AI Assistant.

  • AI-generated responses may be incomplete or contain errors. Progress Software is not responsible for the accuracy or content of AI-generated responses.

Tips on Using the AI Assistant

AI systems are highly effective at interpreting natural language requests, but here are several tips to help you get started:

  • Copy and use the response: The AI Assistant's response window is an HTML browser widget, offering familiar behaviors from popular web browsers. Most notably, you can select and copy text from the response to paste it elsewhere, including into your rule assets.
  • Be descriptive in your AI prompt: The more detailed and specific your question or request, the higher the likelihood that the AI-generated response will meet your needs.
  • Try the prompt again: AI-generated responses may vary even if the question or request remains the same. If the response doesn’t meet your expectations, try submitting the prompt again.
  • Ask for HTML format: The AI Assistant requests responses in HTML format by default. If the AI doesn’t comply, try resubmitting the prompt or include instructions like "Display as HTML" in your request.
  • Request different response formats: The AI can produce responses in various formats. For example, including instructions such as "Output as JSON" can generate a response in a format that better suits your needs.
  • Ask why: If you don’t understand a particular AI-generated response, you can ask the AI for an explanation. Including directions like "Explain why" or "Explain your reasoning" often provides deeper insights into your rule assets.
  • Experiment with different prompts: Feel free to explore and try different types of questions or requests to discover what works best for your specific needs.

AI Assistant is Language Aware

The AI will recognize the language used in prompts you enter. Enter your question or request in any language supported by OpenAI, for example:

The generated response will match the language: