Communication Intelligence Lab

Reclaiming Senior Time
Without Losing Your Voice

Executive communication is a leverage point: critical emails, difficult conversations, and application positioning consume senior time and carry relationship risk. This demo shows AI as a reviewable communication copilot, not an autonomous sender.

1Choose
2Review
3Generate

Step 1

Choose a starting point

Start with a realistic executive email scenario or begin from a blank thread.

Step 2 lets you edit the thread, goal, and audience before generating.
Under the hood: architecture, operating model, and implementation pathSupporting detail is collapsed so the main demo stays focused on business value and the active workflow.

Capability proof

Multi-agent communication workspace

Service model

Multi-agent workspace with distinct input and output contracts.

Intelligence layer

Drafts replies, coaches difficult messages, and positions job applications.

Operational state

Keeps each agent tied to a scenario, audience, goal, and structured result.

Human control

Outputs are designed for review and refinement, not automatic sending.

Business value

Compresses senior communication work while preserving judgment and voice.

System Architecture, Current Build

Agent Modules

email-studio-agentClaude claude-sonnet-4-20250514Thread → tone assessment + draft + key moves + watch-out
text-coach-agentClaude claude-sonnet-4-20250514Conversation + goal → diagnosis + rewrite + next move
resume-fit-agentClaude claude-sonnet-4-20250514JD + resume → fit score + strengths + gaps + positioning

Frontend

communication-intelligence-labReact + TypeScriptVite SPA · tabbed agent workspace · domain-typed API layer
emailStudioApi.tsAPI clientStructured output contract · local fallback via emailLogic.ts
resumeAgentApi.ts / textCoachApi.tsAPI clientsSame pattern, VITE_EDGE_API_BASE toggle for live vs local mode

Domain Layer

emailLogic.ts / emailTypes.tsTypeScriptDeterministic scoring for offline use, swapped by API when backend present
resumeLogic.ts / textCoachLogic.tsTypeScriptSame pattern, stable domain contracts decouple UI from backend

Roadmap

1Deploy FastAPI backend (Python) behind API Gateway for live scoring
2Add EWS mail fetch integration, pull real inbox threads for triage
3Persist draft history in DynamoDB for cross-session continuity
4Add voice-to-text input for meeting transcript coaching
5Fine-tune routing model on communication style preferences

Current Maturity

Three live AI agents running on Claude claude-sonnet-4-20250514 with structured JSON output contracts. Frontend is TypeScript + React with domain-typed API layer and local fallback mode. Backend (FastAPI) is designed but not yet deployed.

Communication Intelligence Lab · Live structured outputView source on GitHub →