Presentation
Inspiration
Estimating the time and effort required to implement business requirements in software projects is a tedious, time-consuming process. Most organizations have vast amounts of historical data — Jira tickets, previous estimations, and business analysis documents — but this knowledge is rarely reused effectively. We wanted to change that by creating an AI-powered agent that not only understands new business requirements but can intelligently reference past work to generate accurate, explainable estimations.
What it does
Our agent, built in Google Agentspace, takes as input a PDF file containing business requirements for a new software implementation project. Leveraging connected data sources — a Google Drive folder with historical business analysis and estimation documents, and a Jira instance containing tens of thousands of development and testing tickets — the agent performs the following:
Understands the PDF content using AI-driven document comprehension.
Identifies relevant past projects and Jira tickets with similar scopes.
Generates a structured list of tasks required to implement the new project.
Estimates effort (in hours) for each task based on historical data and reasoning.
Explains each estimation with a justification and cites the source of inspiration (similar Jira ticket or previous project document).
The output is a clean, AI-generated table with the following columns:
Ticket Title
Ticket Description
Estimation (hours)
Reason for Estimation
Source of Estimation
How we built it
We used Google Agentspace to create a custom agent with access to two connected data stores:
A Google Drive folder containing structured and unstructured documents from past projects (business specs, estimates, testing activities).
A Jira instance with thousands of tickets and associated effort estimations.
The agent is designed to:
Parse and understand the PDF document using Google's built-in AI capabilities.
Query the connected data stores for related projects or tickets.
Synthesize new tasks based on similar historical patterns.
Output results in a structured table format directly within the chat interface.
We fine-tuned prompts to ensure high relevance in estimations and traceability back to data sources.
Challenges we ran into
Document understanding: Parsing diverse PDF structures and extracting meaningful content reliably.
Similarity mapping: Matching current business requirements with relevant Jira tickets or Drive files was complex and required iterative testing.
Explainability: Ensuring that each estimation came with a clear and verifiable reason and source added an additional layer of logic to the system.
Accomplishments that we're proud of
Developed a fully working agent capable of reading business requirements and generating explainable estimations in minutes.
Successfully connected and leveraged real-world data sources (Google Drive + Jira) to support AI decision-making.
Achieved context-aware estimations by anchoring them in historical, project-specific evidence.
What we learned
Agentspace provides a powerful framework for building intelligent assistants that can operate over complex, multi-source business data.
High-quality estimations require not just AI, but also meaningful context — and the ability to reference past data makes all the difference.
Explainability is key when AI is used in project management decisions — users need to trust the process behind each recommendation.
What's next for Top AI
Improve similarity scoring between input requirements and historical data for better relevance.
Add natural language querying so business users can ask questions like “How long would this take based on past projects?”
Export to Jira: Allow direct ticket creation from the output table.
Team-level calibration: Adjust estimations based on the specific team's historical delivery speed.
Multi-language support: Enable input and data processing in multiple languages for broader adoption.
Top AI is just the beginning — we're building toward a smarter, faster, and more explainable way to estimate and plan software projects.
Built With
- agentspace
- custompromptengineering
- drive
- gemeni
- google-cloud
- jira
- pdfparser
Log in or sign up for Devpost to join the conversation.