Feature Requests

Request for More Granular Magic Title Controls
TLDR: Magic Titles only support one global 5,000 character prompt, which is too limited for both general rules and the conditional logic needed for deterministic tickets. Requesting options like per board prompts, conditional prompts, or flow level prompts to improve accuracy and control. --- Hi team, After reviewing the Magic Title Rule Creation Guide and the configuration options, I wanted to share feedback on a limitation that is affecting how we automate our ticket titling process. Magic Title currently uses a single global prompt for all boards and all tickets, with a 5,000 character limit. This single prompt does not give us enough control or space to handle both general title logic and the more specific conditional rules needed for certain types of tickets. We get many deterministic tickets that always use the same wording, especially security alerts. For these, we want to eliminate guesswork and provide exact conditional instructions. For example, if the ticket contains specific keywords, then we want Magic Title to apply a specific title format. Because the global prompt must cover everything at once, we do not have enough room to define both broad rules and detailed conditional logic. Based on your documentation, it looks like Magic Titles are meant to be simple, consistent, and easy to configure, but not yet designed for deeper branching logic. Since I am not sure what direction you want Magic Titles to move toward, I wanted to ask for guidance and share some potential solutions that would solve the challenge. Possible approaches: • Per board Magic Title prompts Each board could have its own smaller, more focused prompt without hitting the character limit. • Multiple title prompts applied conditionally Allow separate title rule blocks that run only when certain criteria are met, such as specific keywords, flow triggers, or metadata. • Flow level title prompts Allow flows to pass a specific Magic Title prompt instead of using the single global prompt. This would make the system similar to how priority prompts work. • Increased prompt size If granular controls are not planned, increasing the global prompt limit would help, although it is more of a temporary workaround. Why this matters: We want consistent, predictable, and accurate titles. Your documentation emphasizes clarity and standardization, but without more granular control, Magic Title has to be kept extremely general. That makes it difficult to handle deterministic ticket patterns that should not rely on guessing. Before we design our own workarounds, I want to make sure we are using Magic Titles as intended. If the feature is meant to stay simple and global, advice on handling deterministic patterns within that structure would be helpful. If more granular control is planned, any insight into that direction would be appreciated. Thanks for taking the time to consider this.
0
·
Assistive AI
Improve AI Auto-Title Generation by Adding Image/OCR Parsing and Promotional Content Filtering
When tickets are created from image-heavy emails (e.g., Microsoft promotional footers with alt-text like "Get Outlook for iOS"), the AI title generator reads image metadata rather than actual message intent. This results in misleading titles that impact triage, dispatch, and SLA prioritization. Real-world example: a contact reporting "My computer might be hacked" received an AI-generated title of "Install Outlook for iOS" due to a Microsoft promotional footer in the email body, causing the ticket to be triaged as a routine software install rather than a security incident. The request covers five improvements: Image/OCR Parsing - Analyze embedded images for actual visual content, not just alt-text or filename metadata. Promotional Content Filtering - Identify and deprioritize common marketing/promotional image patterns (e.g. "Get Outlook for iOS", "Download our app") when generating titles and summaries. Subject Line Weighting - Weight the email subject line more heavily than image alt-text in the title generation model. Contextual Intent Scoring - When the body text is sparse or image-dominated, fall back to the email subject line as the primary title source. Confidence Flagging - If the AI cannot confidently determine ticket intent, flag the auto-generated title as low-confidence so dispatchers know to manually review before routing. Business impact includes triage accuracy, SLA risk (security incidents misrouted as software installs), client trust in sensitive industries (healthcare, finance), and tech efficiency. Submitted by Craig Wilde, Acrisure Cyber Service.
0
·
Assistive AI
Load More