Qubitz Workflow
The end-to-end pipeline from use case discovery to deployed, managed production-grade AI agents.
The end-to-end pipeline from use case discovery to deployed, managed production-grade AI agents.
1. Use Case Intelligence
Multi-agentic research engine identifies, evaluates, and prioritises AI use cases by business value and feasibility. The system crawls your company website, analyzes uploaded documents, and gathers market intelligence to produce prioritized, actionable AI opportunities.
Results land on the Jamming Board -- a collaborative workspace where your team rates use cases with star ratings, discusses them via threaded comments, and ranks them automatically using combined rating and sentiment analysis. The top-ranked use case gets a "Best Fit" badge.
Artifacts generated during discovery include PowerPoint decks (marketing, technical, strategy, use case), AWS architecture diagrams, and deep research reports.
2. System Architecture & Design
From the Jamming Board, selecting "Design App" on a use case creates an application and opens the Control Hub. The Agentic System tab provides guided multi-agent system design with configurable agent roles, orchestration patterns, models, and data flows.
Domain experts refine, validate, and optimise the agentic system through natural language modifications -- adjusting agent behaviors, models, tool configurations, guardrails, and knowledge base connections in real time before any code is generated.
Alternatively, select a Blueprint from the Blueprints page to skip design entirely. Blueprints are pre-built systems for common use cases (Customer Support, IDP, AI Voice Screening, Hyperpersonalisation) that are customized to your company through a short setup wizard.
3. Application Feature Testbed
Users test their designed agent system in a live testbed environment connected to the actual AgentCore runtime. Full control over agent configurations, tool integrations, and knowledge base connections.
Test conversations validate agent behavior, handoffs, and orchestration patterns. The Agent Activity Tree shows real-time sub-agent orchestration. Changes to prompts, models, or tools take effect immediately without redeployment.
Evaluation metrics are available across Response Quality, Task Completion, Tool Usage, and Safety dimensions.
4. Full-Stack Deployment
One-click deployment provisions infrastructure across Amazon Bedrock AgentCore, ECS Fargate, DynamoDB, S3, Cognito, and CloudWatch. The application deployer agent automatically generates and deploys the full stack -- API endpoints (WebSocket + HTTP), authentication, memory systems, monitoring, and security guardrails.
Your agents go live with built-in conversational memory, real-time health monitoring, and secure API access.
5. Application Management
After deployment, the Application Management dashboard provides ongoing control across ten tabs:
- Overview -- App status, endpoints, and document generation (WAFR Report, Application Design Document)
- Agents -- View and manage deployed user and platform agents
- App Config -- GitHub collaborator, custom domains, API endpoint, channel integrations, and build history
- Test Bed -- Live chat with deployed agents, agent activity tree, and evaluations
- App Refiner -- AI-assisted changes to your deployed application in plain English
- Identity & Access -- User access configuration and authentication settings
- Observability -- Sessions, traces, latency, errors, and token usage
- Compliance & Security -- Security pen tests and DPO Pack generation
- FinOps -- Daily cost trends and usage breakdown by AWS service
- Settings -- Application configuration, access control, and teardown