Platform architecture
IBM Concert platform architecture consists of four layers that work together to create a unified operational system. Each layer serves a distinct role while remaining tightly integrated with the others.
The data access layer collects signals from your environment, core capabilities process and prepare data for use. The agentic AI layer orchestrates cross-capability workflows, and the visualization layer surfaces results and actions.
Data access layer
The data access layer continuously collects signals from applications, infrastructure, networks, and security tools across your environment. Concert platform uses a federated model where each capability collects raw data through its ingestion methods. The layer captures both the signals themselves and the relationships between systems, creating a connected view that spans domains.
This unified data model makes cross-capability queries possible. When you ask a question that spans observability, optimization, and security, the platform can answer it because all capabilities read from and write to the same substrate. The relationships between systems are preserved in the model, so the platform understands how a change in one domain affects another.
The data layer is accessible through a standards-based GraphQL API endpoint or through a model context protocol (MCP) server. The Concert assistant uses the data layer to correlate signals across domains. Capabilities use it to understand system-wide context. Custom views in Concert Home pull from it to show data that spans multiple products.
Core capabilities layer
The core capabilities layer processes data and prepares it for use. Core capabilities include Concert Observe (comprising Core Observability and Network Observability), Concert Optimize, Concert Protect, Concert Resilience, and Concert Operate (future). Each capability contributes signals to the data layer and consumes shared context from it.
Concert Observe ingests telemetry from applications, infrastructure, and networks. It writes observability signals to the data layer and reads from it to understand how application performance relates to infrastructure health and network behavior.
Concert Optimize analyzes resource usage and application demand. It writes optimization signals to the data layer and reads from it to understand how infrastructure decisions affect application performance and cost.
Concert Protect identifies vulnerabilities and operational exposures. It writes security and risk signals to the data layer and reads from it to understand how vulnerabilities map to running applications and infrastructure.
Concert Resilience measures reliability across services and dependencies. It writes resilience signals to the data layer and reads from it to understand how incidents, resource constraints, and vulnerabilities affect overall system health.
Each capability offers multiple ways to access the data and tools available within it through GraphQL, MCP, REST, and a coordinating AI agent. The capabilities are not isolated tools. They share a common substrate, which means a question about GPU efficiency can pull from observability data, optimization recommendations, and cost signals in a single query.
Agentic AI layer
The agentic AI layer processes natural language queries, creates plans to execute them, and facilitates execution with human-in-the-loop approval. The Concert assistant is the entry point where you pose questions or request actions. Behind the assistant, a cohort of AI agents runs cross-capability queries, correlates data across domains, and assembles results.
The assistant orchestrates agents that are aligned to capabilities. When you ask a question that requires data from multiple domains, the assistant identifies which agents to involve, coordinates their work, and synthesizes their results. The agents query the data layer and the capabilities through the MCP and GraphQL interfaces, correlate findings, and return structured answers.
A vector database enhances user requests with added context.
This orchestration pattern enables cross-capability workflows. The assistant does not execute tasks itself. It coordinates the agents that have the domain expertise to answer specific questions or perform specific actions. The agents rely on the data layer to access signals and relationships across the environment.
Human-in-the-loop controls are built into this layer. When an action is recommended, you review and approve it before execution. The platform maintains transparency and auditability throughout.
Visualization layer
The visualization layer generates UI dashboards dynamically by using AI agents and provides personalized, contextual dashboards based on user personas. Concert Home is where you interact with the platform. It renders cross-capability results as role-relevant views and provides the interface for action.
Concert Home pulls data from the data layer and from capabilities to build custom views. A view might show application performance metrics from Concert Observe, resource recommendations from Concert Optimize, and vulnerability counts from Concert Protect in a single dashboard. The view is built from data that spans multiple capabilities, unified through the data layer.
The Concert assistant is accessible from Concert Home. You can ask questions, request custom views, or initiate workflows directly from the interface. The assistant coordinates the work behind the scenes and returns the results to Concert Home for display.
How the layers work together
A request moves through the layers and each layer contributes to the result.
Request flow example
You ask the Concert assistant to investigate an incident that affects an AI application. The assistant receives the request and identifies that it requires data from observability, optimization, and security domains.
The assistant coordinates agents that are aligned to those capabilities. The observability-aligned agent queries the data layer for application performance signals and identifies a latency spike. The optimization-aligned agent queries the data layer for GPU allocation data and finds that GPU resources are constrained. The security-aligned agent queries the data layer for vulnerability information and confirms no active exploits.
The agents correlate their findings. The latency spike coincides with GPU constraint. The data layer provides the relationships between the application, the infrastructure, and the GPU resources, so the agents can connect the signals.
The assistant synthesizes the results and returns them to Concert Home. You see a correlated view that explains the incident: the application is experiencing latency because GPU resources are insufficient to meet demand. The assistant recommends scaling GPU allocation and presents the recommendation for approval.
You review the recommendation and approve it. The assistant coordinates execution through Concert Optimize, which adjusts GPU allocation. The action is logged and auditable.
Custom view example
You ask the Concert assistant to create a view that shows virtual machine health across your environment. The assistant identifies that this request requires data from observability (VM performance metrics), optimization (resource utilization and recommendations), and the data layer (VM inventory and relationships).
The assistant coordinates agents to gather the data. The agents query the data layer and the capabilities, pulling performance metrics, utilization data, and optimization recommendations for each VM. The data layer provides the relationships between VMs, applications, and infrastructure, so the view can show how VM health affects application performance.
The assistant assembles the data into a custom view and renders it in Concert Home. You see a dashboard that shows VM performance, resource utilization, and optimization recommendations in a single interface. The view is built from data that spans multiple capabilities, unified through the data layer.
You save the view and share it with your team. Other users with appropriate permissions can access the view and see the same correlated data.