AI-based evaluation condition automation rule

Summary

Introduce a new AI-powered condition type for Automation Rules that evaluates tickets using a configurable LLM prompt.

Description

It would be useful to have a generic “AI Evaluation Condition” within Automation Rules.

This condition would allow administrators to define a System Prompt and optionally select an AI model (to use instead of that configured in global settings), and then evaluate a ticket (e.g. subject + body) to return a boolean result (true/false).

This would enable dynamic, semantic rule conditions that go far beyond regex/keyword matching.

Proposed Behavior

• The condition sends ticket data (Subject + Body) to an AI model
• The admin defines a System Prompt to control behavior
• The model must return a binary output (1/0 or TRUE/FALSE)
• The result is used as the condition outcome

Example Use Cases

✅ Auto-reply gating

• “Should this ticket be answered automatically?”

✅ Ticket routing

• “Is this ticket about billing?”

✅ Spam / low-quality filtering

• “Is this request meaningful?”

Suggested UI Fields

• System Prompt (textarea)
• Model (free text, optional; for example can use here a gpt-5-nano model instead of regular gpt-5.4 that is a lot more expensive)
• Use HTML body (checkbox)
• Standard “NOT” operator

Example Prompt

Decide whether it is appropriate for an AI to automatically respond.

Return:
1 = yes
0 = no

Prefer YES for how-to questions.
Return NO for issues, errors, or requests requiring investigation.

Output only 1 or 0.