Entire FSM flow, from initial input (Idle) through content validation, project selection, ticket generation, review, and finalization:

About the Project

AutoTasker by PIO streamlines the process of creating structured Jira tickets directly from Confluence pages or user-provided content. By leveraging Atlassian Forge, AI-driven insights, and a carefully designed workflow, AutoTasker by PIO reduces manual overhead and ensures teams can focus on meaningful, value-driven work.

Inspiration

The inspiration for AutoTasker by PIO came from the need to streamline task creation and project management in modern workflows. Managing content across Confluence and Jira often involves repetitive, manual processes that consume valuable time. Our goal was to create an intelligent assistant capable of simplifying this process while maintaining accuracy, consistency, compliance with Jira’s required fields, and user oversight.

Prompt Engineering Approach: Instruction-Based Prompting + FSM

A key differentiator of AutoTasker by PIO is the use of Instruction-Based Prompting combined with a Finite State Machine (FSM). This approach ensures that:

Consistency and Reliability: The assistant follows a structured, predefined workflow (FSM) so that users always know the next step.
Precision in Guidance: Instruction-based prompts provide clear instructions at each state, minimizing confusion and errors.
Robust Error Handling: The FSM design anticipates invalid inputs or missing required fields (such as summary, description, labels, and priority in tickets), providing guidance or fallback solutions to the user.

What We Learned

Building AutoTasker by PIO taught us the importance of designing user-centric workflows that balance automation with control. Through extensive research, we learned:

How to integrate and optimize Atlassian Forge tools for enhanced user experiences.
The value of feedback loops to ensure user-generated tasks are precise and actionable.
The critical role of validation and error-handling to maintain seamless workflows.

How We Built It

Technologies: Atlassian Forge, Rovo (AI-powered LLM), TypeScript, Confluence and Jira APIs.
Ticket Creation: Jira API integration to generate structured tickets. The application ensures each ticket meets the minimum required fields—summary, description, labels, and priority—to avoid validation failures.
Prompt Engineering: A well-defined prompt and FSM ensure that the assistant interacts with the user in a guided manner, from initial content ingestion to final ticket creation.
Error and Permission Handling: Comprehensive checks ensure users have the required project permissions before proceeding.

Features

Seamless Content Extraction: Retrieve Confluence content or accept user-provided text.
Actionable Insights with Rovo: The AI model identifies tasks and transforms them into tickets.
Structured Ticket Creation: Generate Jira tickets with all mandatory fields—summary, description, labels, and priority—ensuring no validation errors occur.
Guided Modification and Review: Users can confirm, modify, add, or remove tickets before finalizing.
Embedded Contextualization: Optionally embed created tickets back into Confluence pages for quick reference.

Challenges We Faced

While developing AutoTasker by PIO, we encountered several challenges:

Content Parsing: Handling diverse content types and reliably extracting actionable items required iterative refining of parsing logic and LLM prompts.
Permission Handling: Validating user permissions dynamically for Jira project selection required robust error-handling mechanisms.
Balancing Automation and Control: Balancing automation with user control during task review and modification was a critical design consideration.
Workflow Complexity: Implementing the FSM required careful planning and iteration to maintain clarity and fluid navigation through states.

Despite these challenges, we refined our approach and created a solution that enhances productivity and simplifies task management.

Potential Errors and Requirements

If the chosen Jira project does not support or require certain fields, tickets must still include summary, description, labels, and priority to avoid validation errors during creation.
The assistant guides the user to correct any missing mandatory fields before finalizing.

What's Next?

Moving forward, we plan to:

Multi-language Support:Add support for multi-language content parsing.
Enhanced AI Extraction: Improving LLM capabilities for more context-aware, accurate ticket generation.
Customizable Templates: Allowing teams to define templates for ticket fields, ensuring even smoother workflows.

Impact and Outlook

AutoTasker by PIO aims to empower teams by eliminating tedious steps and streamlining the journey from idea to actionable Jira tickets. By continuously refining its AI-driven extraction, expanding feature sets, and maintaining a user-focused design, we aspire to become an indispensable tool for teams worldwide.