AI Tasks

Create classification, summary, field extraction, reply draft, and title rewrite tasks.

Agents

An Agent configures an LLM and optional tools (LD-046). Ticket Agents instruct local Runtime automatically; chatbot Agents power the chat widget. Kinds:

KindPurpose
Queue classificationPredict ticket queue
Priority classificationPredict ticket priority
Type classificationPredict ticket type
Service classificationPredict ticket service
Dynamic field classificationPredict dropdown DynamicField value
SummarizationWrite a summary into a field
Field extractionExtract a value from the first message into a DynamicField
Reply draftDraft subject and body for review (optional search_faq)
Title rewriteReplace unclear subjects when above confidence threshold
ChatbotInteractive chat using search_faq and create_ticket

Studio stores Agents locally; Runtime runs active ticket Agents. No training required — classification uses zero-shot from value descriptions (see Classification tasks).

Agents list in Studio

Open the list

  1. Sign in to Studio (Signing in).
  2. Click Agents in navigation.

Each row displays name, kind, target attribute, LLM size, status (Active / Inactive), and optional Archived tag.

Create an Agent

  1. Click + New Agent and select a kind, or pick Try an example.
Menu itemEffect
Queue classificationOpens new queue classification Agent
Priority classificationOpens new priority classification Agent
Type classificationOpens new type classification Agent
Service classificationOpens new service classification Agent
Dynamic field classificationNew DynamicField Agent — set target field name
SummarizationOpens new summarization Agent
Field extractionNew field extraction Agent — set target field, prompt, and type
Reply draftNew reply draft Agent — set prompt, model, tools, and trigger conditions
Title rewriteNew title rewrite Agent — set state, strictness, model, and prompt
Example classification: department (under Examples)Demo: dropdown DynamicField OTAIBeispielAbteilung with department cues
Example queue classification: municipalityDemo: municipal queues as DynamicField OTAIBeispielKommuneQueue
Example priority classification: administrationDemo: high/medium/low priorities as DynamicField OTAIBeispielPrioritaet
Example queue classification: universityDemo: university queues as DynamicField OTAIBeispielUniversitaetQueue

Then configure and Activate:

For field extraction, describe what the target DynamicField holds and pick Text, Integer, or Number. In Workflow settings, choose the ticket state (e.g. new) and enable Only run when the target DynamicField is empty. The task parses the subject and first customer message without overwriting existing values.

Archive

  • Archive hides tasks from the main list while keeping database records.
  • Show archived displays archived tasks.
  • Unarchive restores visibility.
  • Archived tasks cannot run.

Active / Inactive controls Runtime execution. Archived only controls Studio visibility.

AI TASKS in the sidebar provides quick links to existing tasks from any page.