Agentic content management

Content Cortex is the next-generation agentic content management architecture that combines enterprise content services with AI capabilities. The architecture provides repository services for capturing, managing, and storing business-related digital assets in multiple repositories (called object stores). It also provides content services that support AI integration through standardized protocols, enabling intelligent content operations while maintaining enterprise-grade security and governance.

Architecture overview

The following diagram illustrates the Content Cortex architecture that shows how the different layers work together to provide agentic content management capabilities. Explanatory information for each layer follows the diagram.

Content Cortex architecture showing repos layer with object stores, services layer with content services, MCP integration layer, AI agent layer with reasoning and traceability, AI foundation with public cloud services, and application layer

Object stores and repositories

Object stores serve as specialized repositories for different business domains and use cases. You can configure each object store to store content in a database, a file system, or a fixed content device such as IBM® Spectrum Protect. You can also use a combination of these storage options.

Object stores provide flexible storage for diverse content needs. An object store can store a wide variety of business-related data and content types. Examples of stored content include the following items:

Commerce and product information such as product catalogs, pricing data, and inventory records
HR policies, benefits documentation, and onboarding workflows
Loan applications, approval workflows, and regulatory documents for lending and compliance
Contracts, invoices, and supplier records for vendor and operations management
Structured or unstructured content such as XML documents, Microsoft Office documents, web pages, photos, voice data, images, process definitions, and templates

Content services

The content services layer provides essential capabilities that enable content management and AI integration across all object stores:

Content storage: Provides multi-device storage, lifecycle management, and governance capabilities that can ensure that the content is stored efficiently and managed according to business policies.
RAG service: Enables Retrieval-Augmented Generation (RAG), a technique that combines AI language models with enterprise content retrieval. The service uses vector search and embeddings to find relevant content based on semantic meaning rather than just keyword matching. This allows AI applications to provide accurate, context-aware responses grounded in your organization's content.
Enhanced Text Extraction service: Provides document processing capabilities such as text extraction from images, handwriting recognition, form processing, and table extraction. This service prepares unstructured content for AI analysis and processing.
Permissions and governance: Manages access control, legal holds, retention policies, and audit capabilities that can ensure content security and compliance with regulatory requirements.

Model Context Protocol integration

The platform supports integration with AI agents and applications through the Model Context Protocol (MCP), an open standard that defines how AI applications communicate with content services. MCP eliminates the need for custom integrations by providing a standardized interface for content operations.

The platform implements MCP through remote servers that expose content services capabilities. These servers use the FastMCP protocol, an efficient implementation designed for high-performance agent-to-service communication:

Core capabilities: The Core MCP server enables AI agents to discover and access content programmatically. It provides document search, folder navigation, content retrieval, and metadata queries.
Classification capabilities: The Classification MCP server helps organize and enrich content automatically. It provides auto-classification, property extraction, and metadata-mapping services.

MCP integration allows various AI clients (such as IBM watsonx Orchestrate, ChatGPT, Claude, and custom applications) to interact with content services by using standardized tools, resources, and prompts.

AI agent integration

The platform supports AI agent integration through a specialized layer that enables intelligent content operations. This layer is built on industry-standard technologies: LangGraph (a framework for building stateful AI agents), FastAPI (a modern web framework for building APIs), and MCP client libraries. Together, these technologies enable AI applications to perform complex content operations:

Reasoning service: Enables task planning, tool selection, and policy enforcement to help AI agents make intelligent decisions about content processing and workflows.
Agent capabilities: Provides content discovery, document understanding, and intelligent retrieval services that allow applications to interact with content through natural language interfaces.
Traceability: Maintains source citations and question-and-answer logs to ensure transparency and auditability of AI-driven content interactions.

Important: AI-generated content and suggestions should be reviewed and verified before use. AI agents may produce inaccurate or incomplete results. Always validate AI-generated classifications, property values, and content analysis before applying them to production systems.

AI foundation

The platform can integrate with public cloud AI services from watsonx.ai and Microsoft Azure Foundry to provide AI capabilities:

Large language models (such as Granite, Llama, Mistral, and GPT models) for natural language understanding and generation
Embedding models (such as slate and granite-multilingual) for semantic search and vector-based content retrieval
Agent execution environments for running AI-powered workflows
Context reasoning capabilities for understanding content relationships and meaning
Policy enforcement mechanisms to ensure AI operations comply with business rules

Business value

Organizations face challenges in extracting value from vast amounts of unstructured content spread across multiple repositories. Traditional content management systems require manual processes for content discovery, classification, and retrieval. This architecture addresses these challenges by combining enterprise content management with AI capabilities.

Key business benefits include:

Accelerated content discovery through semantic search and AI-powered retrieval
Reduced manual effort through automated classification and metadata extraction
Improved decision-making with AI agents that can reason over enterprise content
Standardized AI integration through open protocols that reduce custom development costs
Enterprise-grade security and governance that maintains compliance while enabling AI innovation

Use cases

The architecture supports diverse business scenarios across industries:

Financial services: Banks use AI agents to analyze loan applications by retrieving relevant documents from compliance repositories, extracting key information, and providing recommendations based on regulatory requirements. The MCP integration allows loan officers to interact with content through natural language queries.
Healthcare: Healthcare providers deploy AI agents that can search across patient records, medical images, and clinical documentation to support diagnosis and treatment planning. The Enhanced Text Extraction service processes handwritten notes and forms, while governance controls ensure HIPAA compliance.
Manufacturing: Manufacturers use AI agents to access product specifications, quality control documents, and supplier contracts. The RAG service enables engineers to find relevant technical documentation through semantic search, reducing time spent searching for information.
Legal and compliance: Legal teams deploy AI agents that can analyze contracts, regulatory filings, and case law stored in object stores. The traceability features maintain audit trails of AI-driven content analysis, supporting compliance requirements.