An AI-Powered Bug Reporting from Screen to Structured Jira Tickets
The solution serves QA and engineering teams at mid-to-large B2B SaaS companies operating in fast-paced development environments. These organizations typically manage dozens to hundreds of test runs daily across distributed teams, where inefficient bug reporting creates bottlenecks in hotfix deployment and release cycles.
Why AWS
Amazon Bedrock provides managed access to frontier models (Claude 3.5) with enterprise controls, while Amazon Textract reliably extracts on-screen text for grounding. Using Amazon S3, Amazon EFS, and Amazon ECS Fargate keeps ingest and processing close to storage, and Amazon EventBridge coordinates processing steps without custom schedulers. This combination reduced undifferentiated operations and accelerated production readiness.
The Challenge
Quality Assurance teams frequently rely on screen recordings to report bugs. However, manually converting those recordings into structured Jira tickets is tedious, inconsistent, and time-consuming—especially in fast-paced product environments where developers need rapid, accurate feedback.
Traditional bug reporting methods delay resolution, lead to communication breakdowns, and reduce engineering efficiency.
The platform was created to eliminate these bottlenecks by automatically transforming screen recordings into detailed, structured Jira tickets using AI.
The Platform Solution
This Platform is an AI-powered bug reporting platform that automates the entire process, from video upload to Jira ticket creation. It supports both real-time screen recording via a Chrome Extension and uploading pre-recorded videos via the web.
Once uploaded, the platform extracts UI content and audio from the recording, processes it through Claude 3.5 via Amazon Bedrock, and generates high-quality Jira tickets complete with reproduction steps, severity, environment details, and more.
Technologies Used
- Frontend: React (Web App & Chrome Extension)
- Backend: FastAPI, Python
- Cloud Infrastructure: Amazon S3, Amazon EFS, Amazon ECS (Fargate), Amazon RDS (PostgreSQL)
- AI Stack: Amazon Textract, Amazon Bedrock (Claude 3.5)
- Integrations: Jira REST API, AWS EventBridge
How It Works
Web Upload Flow
- User uploads a pre-recorded screen recording via the platform’s web app
- Video is stored in Amazon S3
- Amazon Textract extracts on-screen text from frames. Amazon Transcribe generates an audio transcript when a microphone capture is present, and both are used to ground the generated ticket content.
- Claude 3.5 (via Bedrock) processes the data and generates a structured Jira ticket.
- The ticket is automatically published to the relevant Jira project
Chrome Extension Upload Flow
- User starts screen recording via the platform's Chrome Extension
- Every second, a video chunk is uploaded to the backend and stored in Amazon EFS
- After recording, chunks are combined server-side and uploaded to Amazon S3
- The same processing flow as Web Upload is triggered (Textract ➜ Bedrock ➜ Jira)
Key Features
- Multi-Channel Uploads – Web or live Chrome Extension
- Scalable Chunk-Based Uploads – Real-time, server-side stitching via EFS
- Claude 3.5 + Textract Powered – Advanced bug understanding and transcription
- Structured Jira Tickets – Auto-filled fields: steps, severity, description
- AWS-Native Architecture – Secure, reliable, and scalable
Results & Impact
- Bug report creation time: 12–15 minutes → ~3 minutes per issue (~80% reduction).
- Ticket completeness: +45% increase in tickets that include steps to reproduce, environment, and severity (sample n = 200).
- Engineer response start time: ~35% faster from ticket creation to first developer comment.
- Standardization: Consistent ticket structure reduced back-and-forth triage across teams.
Measurement:
Time is measured from upload/record stop to Jira ticket creation using app telemetry. Completeness is scored against a rubric on a weekly sample. Response time is computed from Jira events over a rolling 28-day window.
Security & responsible use
All data in transit uses TLS, and data at rest in Amazon S3/EFS/RDS is encrypted with AWS KMS. Access is granted via least-privilege IAM with audit trails enabled through AWS CloudTrail. Prompts and outputs are designed to prevent the insertion of secrets into tickets, and model decisions are logged with human review controls available for sensitive projects.
What Users Say
“With the new platform, we just record and move on. The AI handles everything else, and the ticket quality is better than what we used to write by hand.”
— QA Manager, B2B SaaS Company
What’s Next for The Platform
- Visual bug mapping with screen interaction overlays
- Multi-language transcript analysis
- Slack & Teams notifications for created tickets
- Enterprise-grade RBAC and audit trails
Our Partners
By using this site, you agree to thePrivacy Policy.