Report 4.1.4 — Smart Priority Tagging Detailed Guide
This document explains the smart priority-tagging feature that assigns HIGH/MEDIUM/LOW based on urgency cues inside the transcript.
Mechanics of Priority Tagging
- The transcript is forwarded to
ActionItemExtractorService.
- A prompt emphasises urgency keywords (e.g., “ASAP”, “urgent”, “by EOD”, “critical”) alongside standard fields.
- The OpenAI completion replies with structured action items containing the preset
priority field.
- The extractor normalizes priorities to three canonical values and uppercases them before handing off to HubSpotTaskService.
Keyword examples
HIGH: “as soon as possible”, “immediately”, “critical”.MEDIUM: “by tomorrow”, “soon”, “follow up”.LOW: “eventually”, “sometime next week”, “whenever you have time”.
Tuning & Observability
Track priority accuracy by examining:
- Action item
priority field before insertion into HubSpot.
- HubSpotTaskCreationReport, which contains the raw responses so you can verify whether the AI interpreted the urgency correctly.
Iterate on prompt wording to adjust how keywords influence priority results.
Fallbacks
- If OpenAI omits the priority field, the extractor defaults to
NONE (mapped to LOW) so tasks still create.
- Priority normalization ensures HubSpot always receives HIGH/MEDIUM/LOW even when the LLM produces synonyms.