# Treasure Data Product Knowledge Base: AI Marketing Cloud & Super Agent

**Document Version:** 2025 (Post-CDP World)
**Status:** Latest Strategic Architecture & Vision
**Scope:** Core Platform, AI Agent Foundry, Super Agent, Native Delivery (ESP/CEP)

## 1. Executive Summary: From CDP to AI Marketing Cloud

In 2025, Treasure Data evolved from a traditional **Customer Data Platform (CDP)** to the **"AI Marketing Cloud."**
This represents a fundamental shift from a static "System of Record" to an autonomous "System of Intelligence and Action."

### Core Concept: Agentic Marketing

A marketing paradigm where humans define the **Goals** (e.g., "Increase retention by 5%"), and AI Agents autonomously orchestrate tools, analyze data, and execute campaigns to achieve them.

## 2. Technical Architecture (The Stack)

The platform is built on four distinct layers. The AI cannot function without the data foundation below it.

### Layer 1: Intelligent CDP (The Foundation)

The "Engine Room." Before AI can act, data must be ingested, unified, and managed at scale.

#### 1.1 Core CDP Functionality (The Data Pipeline)

1. **Collect (Ingestion):**
  * Aggregates data from offline (POS, CRM) and online (Web, Mobile, IoT) sources.
  * **Schema-on-Read:** Ingests unstructured or semi-structured data without rigid pre-modeling.
  * **Integrations:** 100+ pre-built input connectors.
2. **Data Workbench:**
  * A tool for managing databases, creating tables, running queries (Trino/Hive), and managing workflows to transform and prepare data.
3. **Unify (ID Resolution):**
  * **Identity Stitching:** Deterministic and probabilistic linking of fragmented identifiers (cookies, emails, device IDs).
  * **Canonical ID:** A persistent, unique identifier for every customer profile.
4. **Analyze (Segmentation & ML):**
  * **Audience Studio:** Rule-based segmentation interface.
  * **Predictive Scoring:** Native ML for churn risk, purchase propensity, and LTV prediction.
5. **Activate (Syndication):**
  * Connectors to push segments/attributes to external channels (Ad platforms, CRMs).


#### 1.2 The "Diamond Record"

* **Definition:** The output of the "Unify" stage. A clean, trustworthy, legal-ready customer profile.
* **Role in AI:** Acts as the **"Grounding"** source. AI agents query this record to prevent hallucinations.


### Layer 2: AI Agent Foundry (The Engine)

The governance and execution environment for AI.

* **Tech Stack:** Amazon Bedrock (Claude 3, Titan, Llama) + RAG (Retrieval-Augmented Generation).
* **Key Differentiator:** Native access to the Diamond Record allows agents to reason with full customer context.


### Layer 3: Marketing Super Agent (The Orchestrator)

* **Definition:** A "Manager" AI that oversees "Small (Specialized) Agents."
* **Capabilities:**
  * **Intent Interpretation:** Decomposes abstract goals into tasks.
  * **Delegation:** Assigns tasks to specialized agents (e.g., "Audience Agent, find the users," "Creative Agent, write the email").
  * **Swarm Intelligence:** Coordinates multiple agents to complete complex workflows.


## 3. Key Capability: Journey Orchestration

Journey Orchestration transforms static customer data into dynamic, ongoing experiences. It moves beyond "Batch & Blast" emails to real-time, state-based interactions.

### 3.1 Engage Studio

The central tool within the **Engagement AI Suite** for mapping and executing customer journeys.

* **Visual Canvas:** A drag-and-drop interface to design multi-step, multi-channel flows.
* **State-Based Logic:** Users move through the journey based on real-time triggers (e.g., "Wait 2 days," "If email opened, send X," "If purchase made, exit journey").


### 3.2 AI-Driven Optimization

* **Journey Optimizer Agent:** Analyzes active journeys to identify bottlenecks (drop-off points) and suggests structural improvements.
* **Next Best Action (NBA):** AI determines the optimal branch or content for a user at any given node in the journey.


## 4. Detailed AI Suites

### 4.1 Engagement AI Suite (Lifecycle Marketing & Native Delivery)

This suite allows Treasure Data to function as a full **Customer Engagement Platform (CEP)** and **ESP**, replacing the need for external marketing automation tools.

* **Native Channel Delivery (Built-in ESP/CEP):**
  * **Direct Sending:** TD can now directly send messages without relying on 3rd party integrations (like Mailchimp, Braze, or SFMC).
  * **Supported Channels:**
    * **Email:** High-volume email sending with built-in deliverability management.
    * **Mobile Push:** Native push notifications for iOS and Android apps.
    * **In-App Messaging:** Overlays and messages inside mobile applications.
    * **SMS:** Text messaging capabilities.
* **Audience Agent:** "Text-to-SQL" capability. Marketers describe a target group in natural language, and the agent generates the segment.
* **Real-time Triggers:** Instant execution based on live behavioral signals (e.g., cart abandonment, page view).
* **ROI Reporting Agent:** Uses Multi-Touch Attribution (MTA) to analyze channel performance and answer questions like "Which campaign drove the most revenue?"


### 4.2 Creative AI Suite (Content Production)

* **Generation Agents:** Produces subject lines, body copy, and images tailored to the specific audience segment.
* **Brand Guideline Agent:** The "Guardrail." Reviews all AI-generated content against uploaded brand rules (tone of voice, hex codes, forbidden terms) to ensure safety.
* **Image Remix Agent:** Automatically resizes and reformats visual assets for different channels.


### 4.3 Personalization AI Suite (Web/App Experience)

* **Personalization API:** Delivers decisions to edge devices (Web/App) in milliseconds.
* **Profiles API:** Allows external systems to look up customer attributes in real-time.


## 5. Important Distinctions & Terminology

### "Marketing Super Agent" vs. "Marketing Copilot"

| Feature | Marketing Copilot | Marketing Super Agent |
|  --- | --- | --- |
| **Role** | Assistant (Co-pilot) | Manager (Supervisor) |
| **Mode** | Reactive (Waits for user prompts) | Proactive (Autonomous execution) |
| **Scope** | Single Task (e.g., "Help me write SQL") | Full Workflow (Plan → Create → Execute) |


### Market Positioning

* **Vs. Monolithic Clouds:** TD prevents data silos by keeping the CDP as the central brain.
* **Vs. Composable Data Stacks:** TD reduces the "Integration Tax."
* **Transformation:** TD has transitioned from **CDP (Data only)** to **CDP + CEP (Data + Engagement)**. This allows for "No Compute Pricing" models and consolidated vendor stacks.


## 6. Business Value (Why use TD?)

1. **Vendor Consolidation (TCO Reduction):** Companies can retire separate ESPs and Marketing Automation tools, running everything from the CDP data foundation.
2. **3x Faster Time-to-Market:** Agents automate the "wait time" between data extraction, creative production, and approval.
3. **Governance & Trust:** "Zero Hallucination" approach by grounding all AI outputs in the Diamond Record.


## 7. Appendix: Terminology & Directory Mapping

### 7.1 Product Key to Directory Mapping

Use these mappings when navigating the `products/` directory structure.

| Product Name | Directory Key (`products/[key]`) | Notes |
|  --- | --- | --- |
| **AI Agent Foundry** | `customer-data-platform/ai-agent-foundry` | Contains Agents, Foundry Workspace |
| **Audience Studio** | `customer-data-platform/audience-studio` | Segmentation, Activation |
| **Data Workbench** | `customer-data-platform/data-workbench` | Manage databases, create tables, run powerful queries and manage workflows |
| **Integration Hub** | `customer-data-platform/integration-hub` | Connectors |
| **ID Unification** | `customer-data-platform/id-unification` | Identity resolution |
| **Journey Orchestration** | `customer-data-platform/journey-orchestration` | Visual journey builder |
| **Treasure AutoML** | `customer-data-platform/machine-learning/automl` | No-code ML |
| **Precision ML** | `customer-data-platform/machine-learning/precision-ml` | Custom ML (Python) |
| **Treasure Insights** | `customer-data-platform/treasure-insights` | Reporting/Dashboarding |
| **Job Management** | `customer-data-platform/job-management` | Jobs monitoring |
| **Engage Studio** | `marketing-cloud/engage-studio` | Email, Push, SMS (Native Delivery) |
| **Control Panel** | `control-panel` | Admin settings |


### 7.2 Old vs. New Terminology

Be careful to use the **New** terminology in all current documentation.

| Old Term / Concept | New Term / Concept | Context of Change |
|  --- | --- | --- |
| **CDP (Customer Data Platform)** | **AI Marketing Cloud** (Platform Level) | Platform evolution from Data-only to Data + AI + Action. |
| **Email Marketing** | **Native Channel Delivery** | Now includes Email, Push, In-App, SMS natively. |