🔹Azure Architecture Blog·February 12, 2026

Pantone's Agentic AI Architecture with Azure Cosmos DB

This article details Pantone's architectural approach to building an agentic AI-powered Palette Generator using Azure. It highlights the critical role of Azure Cosmos DB as a real-time data layer for managing conversational context, user interactions, and prompt data, emphasizing its scalability and flexibility for AI-driven applications. The architecture incorporates a multi-agent system and is designed to evolve towards vector-based workflows for enhanced semantic understanding.

AI & ML Infrastructure Databases & Storage Distributed Systems

Read original on Azure Architecture Blog

Architecting an Agentic AI System

Pantone developed an "Agentic AI" experience called the Palette Generator. This system aims to provide designers with instant, curated color palettes through a chat-based interface, leveraging decades of Pantone's expertise. The core architectural decision was to implement a multi-agent system, where specialized agents collaborate to fulfill user requests.

Specialized Agents: The system employs agents like a "chief color scientist" and a "palette generation agent." Each agent is responsible for distinct functionalities, such as reasoning, context retrieval, and generating responses.
Dynamic Interaction: Unlike static recommendation systems, this multi-agent architecture responds dynamically to user intent, conversational context, and historical interactions, making the experience highly interactive and personalized.

The Role of an AI-Ready Database: Azure Cosmos DB

A foundational element of Pantone's architecture is Azure Cosmos DB, which serves as the real-time data layer for the Palette Generator. Its selection was based on key requirements for an AI-driven application:

Real-time Data Storage: Stores and manages chat history, prompt data, message collections, and user interaction insights, crucial for responsive, fast, and context-aware agents.
Scalability and Global Distribution: Enables Pantone to serve a global user base with fast data retrieval, handling massive volumes of operational data as AI workflows evolve.
Flexibility: Supports rapid iteration without significant re-architecture, essential in the fast-moving AI landscape. It also supports future enhancements like vector-based workflows, integrating with agent orchestration and embedding models.

💡

AI-Ready Database Characteristics

An "AI-ready" database needs to handle operational data at scale, support conversational memory, enable advanced analytics, and adapt to evolving AI workflows, including vector search capabilities for semantic understanding. This goes beyond traditional transactional requirements.

Evolving Architecture: From Text to Vectors

Pantone's architecture is designed for continuous improvement. A key future step involves transitioning from traditional text storage to vector-based workflows. This involves embedding user prompts and contextual data, enabling vector search, and enriching responses with deeper semantic understanding. Azure Cosmos DB is positioned to support this evolution by handling vectorized data and integrating with embedding models via Microsoft Foundry.

agentic AIAzure Cosmos DBmulti-agent systemvector searchreal-time datascalabilityconversational AI

Comments

Loading comments...

Architecture Design

Design this yourself

Design an agentic AI system for a creative domain (e.g., design, content generation) that leverages a multi-agent architecture, real-time data storage for conversational context and history, and is scalable to support global users. Focus on the data foundation, agent orchestration, and how to evolve the system for vector-based semantic understanding.

Focus: multi-agent AI system with real-time data layer

Other design angles

· Design a multi-agent conversational AI system focusing purely on the interaction between specialized agents and their roles.· Design the data layer for an AI application, detailing how it handles conversational memory, prompt storage, and prepares for vector embeddings and search.· Design a scalable platform for deploying and managing agentic AI services, considering aspects like prompt management, model versioning, and feedback loops for continuous improvement.