Service Task

A service task is a task with a small gear icon in its upper-left corner. It represents work performed by an automated service — a web service call, a script, a scheduled job, an AI agent. No human is in the loop at execution time. The model implicitly says "this is automated; the label is what the automation does".

Service tasks contrast with user tasks (person in the loop) and with plain tasks (unspecified performer). Choosing service over plain is a design decision: you are telling the reader — and the execution engine — that this step is non-human. Engines use that information to wire the step to an implementation: an HTTP endpoint, a serverless function, a Kafka message, an LLM prompt.

• The step is genuinely automated today — a scheduled job, a webhook, an API call, a script that runs on cron.
• The step is automated in the target (AI-transformed) process — you want to show explicitly what the target state looks like when AI takes over this work.
• The step should be automated but is not yet — service task in the "should be" diagram, user task in the "as is". The gap between the two is the automation opportunity.

Service tasks carry the same KPIs as other tasks — estimated duration, cost per execution, frequency — but LucidFlow surfaces them differently. Their duration cost is typically close to zero because software does not bill hourly, but their tool cost (the subscription to the service that runs them) is surfaced separately in the cost dashboard. When the target BPMN replaces a user task with a service task, the ROI report computes the delta between the two automatically: hours saved on the human side, tool cost added on the automation side, net savings per month.