Skip to content

Custom Prompts - Technical Overview

12Port’s AI-Powered Analysis and Insights framework supports the creation of custom AI prompts, enabling organizations to tailor AI behavior and output to specific operational requirements. Custom prompts provide flexibility beyond default system prompts, allowing administrators to define precise analysis objectives, response formats, and processing constraints aligned with internal policies and workflows.

This section describes how to create and manage custom prompts within 12Port and outlines recommended practices for designing prompts that produce consistent and cost-effective AI responses.

Creating Custom Prompts

Custom prompts can be created and managed by users with Administrator or AI Manager privileges. To create a new prompt:

  1. Login with an Administrator or AI Manager account.
  2. Navigate to AI > Prompts and click the Add button.
  3. Enter a unique, but recognizable Name for this prompt.
  4. Define the System Prompt to control AI behavior, formatting, and constraints.
  5. Define the User Prompt to specify the analytical task to be performed.
  6. Toggle on the Enabled parameter.
  7. Click Save to make the prompt available for use.

Best Practices for Prompt Design

When creating custom prompts, consider the following best practices to ensure optimal performance and predictable output:

  • Explicitly instruct the AI to format its output using valid GitHub-flavored Markdown. Output in other formats may not render correctly in the Insights Report response field.
  • Balance prompt complexity with cost and performance considerations. Longer and more complex prompts typically increase processing time and associated AI service costs.
  • Report data is automatically appended to the end of the User Prompt at runtime. While it is not necessary for you to embed the data within the prompt, it is recommended to explicitly note that raw CSV data will be provided as part of the request.
  • Use deterministic language where possible. Clearly specify expected output structure, headings, ordering, and level of detail to reduce variability between AI responses and improve consistency across executions.
  • Scope the analysis explicitly. Define what the AI should and should not analyze (for example, exclude timestamps, user identifiers, or low-severity events) to avoid unnecessary processing and unintended conclusions.
  • Test prompts with representative datasets. Validate new or modified prompts using datasets of varying size and complexity to ensure predictable output and acceptable performance under real-world conditions.

Summary

Custom prompts provide a powerful mechanism for extending AI-Powered Insights within 12Port, enabling organizations to adapt AI analysis to evolving requirements. By following structured prompt design practices and aligning prompt complexity with appropriate AI models, administrators can achieve accurate AI-driven analysis across supported reports.