Dispatch - AI Native Task Coordination

Stop chatting, start doing.

Dispatch is an AI-native task coordination system for B2B teams. Instead of chat-based communication, team members tell the system what they need, and a central AI intelligently routes tasks to the best-qualified available person - or completes the task autonomously if possible.

The Problem

• Messages are unclear between coworkers
• People get assigned tasks when they're already at capacity
• Passive-aggressive communication and workplace friction
• Time wasted on tasks that AI could handle

The Solution

1. Tell the AI what you need - Natural language task requests
2. AI decides the best approach:
   • AI Direct: Completes task immediately (writing, analysis, etc.)
   • Human Assignment: Routes to best-matched team member based on skills & capacity
3. Anonymous task assignment - No bias, just work
4. Automatic reassignment - Tasks get reassigned before deadlines if not on track
5. Smart notifications - Real-time updates on task progress

Tech Stack

• Frontend: Next.js 15 + React + TypeScript + Tailwind CSS
• Backend: Next.js API Routes
• Database: Neon (Serverless Postgres) + Drizzle ORM
• AI: Anthropic Claude API (Sonnet 4.5)

Setup

1. Install Dependencies

bun install

2. Configure Environment

Create .env.local:

cp .env.example .env.local

Edit .env.local:

DATABASE_URL=your_neon_connection_string
ANTHROPIC_API_KEY=your_claude_api_key
NEXT_PUBLIC_APP_URL=http://localhost:3000

Get your credentials:

• Neon: https://neon.tech (free tier, create new project)
• Claude API: https://console.anthropic.com (get API key)

3. Setup Database

bun run db:setup

This creates tables and seeds demo users:

• Sarah Chen (Engineer) - Default user
• Jordan Rivers (Data Analyst)
• Alex Park (PM)

4. Start Development Server

bun run dev

Open http://localhost:3000

Project Structure

dispatch/
├── app/
│   ├── api/              # API routes
│   │   ├── chat/         # AI coordinator endpoint
│   │   ├── tasks/        # Task CRUD
│   │   └── users/        # User management
│   ├── page.tsx          # Home page (TODO: UI)
│   └── layout.tsx        # Root layout
├── lib/
│   ├── db/
│   │   ├── schema.ts     # Database schema
│   │   ├── client.ts     # Neon connection
│   │   └── seed.ts       # Demo data seeder
│   └── services/
│       ├── ai-coordinator.ts    # Claude API integration
│       └── task-assignment.ts   # Assignment algorithm
├── public/
│   └── logo.png          # Dispatch logo
└── drizzle.config.ts     # Drizzle ORM config

API Endpoints

Chat

• POST /api/chat - Send message to AI coordinator
  • Body: { message: string, userId: string }
  • Returns: Task assignment or AI direct result

Tasks

• GET /api/tasks?userId={id}&view={my-tasks|sent|done} - Get tasks
• POST /api/tasks/{id}/complete - Mark task complete
  • Body: { userId: string }

Users

• GET /api/users - List all users

Database Schema

Users

• id (text) - 'sarah', 'jordan', 'alex'
• name, email, role
• skills (array) - e.g., ['code', 'typescript', 'react']
• currentCapacity / maxCapacity - Task load management

Tasks

• id (uuid)
• title, description, deadline, priority
• requiredSkills (array)
• status - 'pending' | 'assigned' | 'in_progress' | 'completed'
• requesterId / assigneeId - Anonymous until completion
• aiCompleted - Boolean for AI direct execution
• aiResult - Result text for AI-completed tasks
• progressPercentage - 0-100

Messages

• Task-scoped chat threads
• Links to tasks for conversation history

How It Works

1. Task Coordination Flow

User types: "I need a summary of last week's metrics by Friday"
        ↓
AI Coordinator analyzes request
        ↓
    Determines execution tier:
    - ai_direct: AI completes immediately
    - human: Assigns to best match
        ↓
Task Assignment Algorithm:
    1. Find users with required skills
    2. Filter by available capacity
    3. Sort by: capacity DESC, skill match DESC
    4. Assign to top match
        ↓
Task created & user notified

2. AI Coordinator

Uses Claude to:

• Parse natural language into structured task data
• Determine if task can be completed autonomously
• Extract title, description, deadline, priority, required skills
• Execute ai_direct tasks immediately

3. Assignment Algorithm

function findBestAssignee(task):
  1. Get all users
  2. Filter: currentCapacity < maxCapacity
  3. Calculate skill match percentage
  4. Filter: has ≥1 required skill
  5. Sort by: available capacity DESC, skill match DESC
  6. Return top match

Development Commands

# Development
bun run dev              # Start dev server

# Database
bun run db:push          # Push schema to database
bun run db:seed          # Seed demo data
bun run db:setup         # Push + seed (fresh start)
bun run db:generate      # Generate migration

# Build
bun run build            # Production build
bun run start            # Start production server

Demo Users

Sarah Chen (Engineer)

• Skills: code, typescript, react, api-design, testing
• ID: sarah

Jordan Rivers (Data Analyst)

• Skills: data, analysis, sql, spreadsheets, visualization, python
• ID: jordan

Alex Park (PM)

• Skills: planning, writing, analysis, spreadsheets, user-research, roadmapping
• ID: alex

Next Steps (Frontend)

• Build main UI (split view: task list + chat)
• Implement task cards with status indicators
• Add real-time updates with WebSocket
• Create task detail view with chat thread
• Add user switcher for demo
• Implement deadline conversation dialog
• Add progress tracking UI

Testing

Test the API directly:

# Get users
curl http://localhost:3000/api/users

# Create task via chat
curl -X POST http://localhost:3000/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "I need help analyzing Q4 metrics by Friday",
    "userId": "sarah"
  }'

# Get my tasks
curl "http://localhost:3000/api/tasks?userId=jordan&view=my-tasks"

# Complete task
curl -X POST http://localhost:3000/api/tasks/{task-id}/complete \
  -H "Content-Type: application/json" \
  -d '{"userId": "jordan"}'

License

MIT