Loading content...
A product concept for permission-aware enterprise knowledge, task and cross-tool workflow agents.
AQL explores an enterprise AI workspace connecting Google Drive, Notion and Jira. Its product architecture separates a Knowledge Agent for permission-aware retrieval, a Task Agent for Jira operations and a Workflow Agent for bounded cross-tool actions. The public repository currently shows an early frontend foundation, so connector execution, permission filtering and BGE reranking must be described as target architecture rather than completed production functionality.
Translating ambiguous multi-stakeholder requirements into deterministic, verifiable software under strict latency, cost, and usability boundaries.
Separate knowledge, task and workflow responsibilities instead of building one unrestricted agent.
Filter access before candidate content reaches generation.
Require typed inputs, audit events and confirmation for material external changes.
Treat the BGE reranker and live connectors as roadmap features until code and evaluation evidence exist.
Test the RAG retrieval pipeline by asking specific questions about this case study: