Mastercard Media Agent Platform (MCMAP)

An enterprise AI agent platform built on Microsoft Power Platform, featuring specialized agents for media planning, marketing strategy, and business consulting.

---

🚀 Quick Start

# Clone repository
git clone https://github.com/kevcofett/Kessel-Digital-Agent-Platform.git

# Choose your deployment branch
git checkout deploy/personal    # For Aragorn AI environment
git checkout deploy/mastercard  # For Mastercard environment

# Follow deployment guide
cat docs/DEPLOYMENT_GUIDE.md

---

📋 Platform Overview

MCMAP consists of 11 specialized AI agents organized into two solution domains:

MPA (Media Planning Agent) Solution

Agent	Code	Description
Orchestrator	ORC	Routes requests to specialist agents
Analytics	ANL	Calculations, projections, statistical analysis
Audience	AUD	Segmentation, targeting, customer value modeling
Channel	CHA	Channel strategy, allocation, media mix
Performance	PRF	Campaign monitoring, optimization, reporting
Supply Path	SPO	Supply path optimization, ad tech strategy
Document	DOC	Media plan generation, presentation creation
Marketing	MKT	Campaign strategy, creative briefs, brand positioning

CA (Consulting Agent) Solution

Agent	Code	Description
Strategy	CST	Strategic frameworks, business analysis
Change Management	CHG	Transformation planning, stakeholder alignment

---

📁 Repository Structure

Kessel-Digital-Agent-Platform/
├── base/                      # Shared base components
│   ├── agents/               # Agent KB files (source of truth)
│   ├── dataverse/            # Schema and seed data
│   └── platform/             # EAP core components
├── release/
│   ├── v6.0/                 # Current release
│   │   ├── agents/           # Agent-specific files
│   │   ├── verticals/        # Industry vertical overlays
│   │   └── docs/             # Release documentation
│   └── v5.5/                 # Previous release (archived)
├── src/
│   ├── adaptive-cards/       # Copilot UI card templates
│   ├── benchmarks/           # Real-time benchmark connectors
│   ├── decision-tree-ui/     # Visual workflow components
│   └── ml-training/          # ML model training pipelines
├── deploy/                   # Deployment scripts
├── docs/                     # Platform documentation
└── tests/                    # Test suites

---

🔧 Technology Stack

Layer	Technology
AI Orchestration	Microsoft Copilot Studio
Data	Dataverse
Automation	Power Automate
Compute	Azure Functions
Knowledge	SharePoint
ML	Azure Machine Learning

---

📚 Documentation

Document	Description
Architecture	System design and patterns
Deployment Guide	Step-by-step deployment
Agent Reference	Agent capabilities and KB inventory
API Reference	Azure Functions and connectors
ML Models	Machine learning components

---

🏭 Deployment Environments

Environment	Branch	Description
Personal	deploy/personal	Aragorn AI (Mastercard)
Corporate	deploy/mastercard	Mastercard deployment
Development	main	Source of truth

---

🧪 Testing

# Run ML model tests
cd src/ml-training && pytest

# Run benchmark connector tests
cd src/benchmarks && npm test

# Run e2e tests (requires environment)
cd tests/e2e && pytest --env=personal

---

📊 Agent Capabilities

ANL (Analytics Agent)

• ROAS, CAC, LTV calculations
• Marketing Mix Modeling (MMM)
• Incrementality testing
• Budget optimization
• Scenario modeling

AUD (Audience Agent)

• Customer segmentation
• Propensity scoring
• Lookalike modeling
• Lifetime value prediction
• Churn prediction

CHA (Channel Agent)

• Channel selection
• Media mix optimization
• Platform-specific playbooks
• Cross-channel attribution

PRF (Performance Agent)

• Real-time monitoring
• Anomaly detection
• A/B test analysis
• Performance optimization

---

🏢 Vertical Support

MCMAP includes industry-specific overlays:

Vertical	Compliance	Key Features
Financial Services	Fair lending, GLBA	Risk-aware targeting, compliance guardrails
Healthcare	HIPAA, PHI protection	Privacy-first measurement, compliant audiences
B2B	Account-based	ABM integration, sales alignment
Retail	PCI awareness	Omnichannel attribution, seasonality

---

📈 ML Models

Model	Purpose	Training Data
Churn Predictor	Customer retention risk	Customer behavior signals
Media Mix Model	Channel contribution	Spend and outcome time series
Lookalike Model	Audience expansion	Seed audience profiles
Response Curve	Diminishing returns	Spend vs. outcome data
Budget Optimizer	Allocation optimization	Multi-channel scenarios
Propensity Model	Conversion likelihood	User features
Anomaly Detector	Performance alerts	Time series metrics

---

🔐 Security

• Azure AD authentication
• Role-based access control
• Data isolation by session
• No PII in knowledge base
• Audit logging

---

📄 License

Proprietary - Mastercard © 2024-2026

---

🤝 Contributing

1. Create feature branch from main
2. Follow 6-Rule Compliance for Copilot docs
3. Run validation suite
4. Submit PR with test results

---

📞 Support

• Technical Issues: Create GitHub issue
• Deployment Help: See Deployment Guide
• Architecture Questions: See Platform Architecture