Generative AI capabilities in Instana

Edit online

Instana uses generative AI and large language models (LLMs) to enhance observability, accelerate incident resolution, and automate troubleshooting workflows. These AI-powered capabilities help DevOps and SRE teams reduce mean time to resolution (MTTR), improve operational efficiency, and gain deeper insights into complex distributed systems.

AI gateway

The AI gateway is the foundational infrastructure that powers all AI capabilities in Instana. It manages connections to watsonx.ai or other external LLM services through configurable LLM gateways, enabling you to control which models and services power each AI capability.

Key features of the AI gateway:

  • Centralized management of LLM connections and configurations
  • Support for multiple LLM services (watsonx.ai, vLLM-compatible services)
  • Capability-specific gateway assignments for fine-grained control
  • Default gateways for SaaS environments
  • vLLM configurations for self-hosted deployment options

Each AI capability in Instana can be configured to use a specific LLM gateway in SaaS environments.

For more information, see Connecting Instana to an LLM gateway.

AI capabilities

The following AI capabilities are available in Instana:

Incident summarization uses generative AI to create concise summaries of incidents by analyzing impacted entities, related events, affected entities, and incident notes. The AI-generated summaries help teams quickly understand incident context and accelerate root cause identification.

What incident summarization does:

  • Generates comprehensive incident summaries that include start time, duration, severity, and status
  • Summarizes impacted entities, related events, and affected entities
  • Summarizes incident notes, including notes from bidirectional alert channels
  • Identifies the triggering entity by using topology analysis
  • Enables sharing through email with auto-populated content
  • Provides tailored summaries for different audiences such as SRE, DevOps, developer, executive, IT manager, and user

For more information, see Incident summarization.

Intelligent incident investigation combines agentic AI, causal AI analysis, and LLMs to automatically identify probable root causes across your entire environment. This multi-entity investigation capability analyzes system topology, distributed tracing, metrics, logs, and infrastructure events in parallel.

What intelligent incident investigation does:

  • Completes automated root cause analysis (RCA) across services, infrastructure, and Kubernetes resources
  • Conducts multi-entity investigations that analyze entire failure propagation chains
  • Correlates change events, configuration updates, and deployments with incidents
  • Provides evidence-driven recommendations with supporting metrics, logs, and traces
  • Reduces MTTR by 60-80% through automated analysis
  • Generates plain-language investigative guidance for incident response

For more information, see Incident investigation.

AI assistant is a conversational interface that uses natural language to help you analyze and understand observability data. It transforms complex monitoring information into clear, actionable insights through simple conversational queries.

What the AI assistant does:

  • Enables natural language queries across application calls, infrastructure components, and incidents
  • Supports multiple technologies such as Db2, Kubernetes, IBM MQ, and JVM
  • Provides interactive data tables with export capabilities
  • Includes a prompt library with pre-built query templates
  • Uses multi-agent architecture for intelligent query routing
  • Delivers real-time results with transparent query interpretation

For more information, see AI assistant.

Kubernetes AI assistant specializes in Kubernetes cluster troubleshooting, allowing you to ask natural language questions about cluster health, pod issues, namespace resources, and deployment status.

What the Kubernetes AI assistant does:

  • Provides Kubernetes-specific troubleshooting through natural language queries
  • Includes a pre-built prompt library for common Kubernetes scenarios
  • Generates automation scripts and manual actions for event remediation
  • Saves generated actions to the Action Catalog for reuse
  • Integrates with Kubernetes monitoring and observability data

For more information, see Kubernetes AI assistant.

Action generation using AI uses generative AI to create remediation actions. This AI uses retrieval-augmented generation (RAG) patterns to generate manual action descriptions and automation scripts.

What action generation using AI does:

  • Generates next best steps to diagnose and remediate issues and incidents by using RAG patterns
  • Creates automation BASH scripts and Ansible playbooks by using code generation models
  • Supports kubectl commands and other automation frameworks

For more information, see Action generation using AI.

SLO AI assistant provides AI-powered analysis and insights for Service Level Objectives, helping you understand SLO health, identify trends, and receive optimization recommendations.

What the SLO AI assistant does:

  • Generates instant summaries of SLO status and health
  • Identifies root causes and recommended actions during SLO violations
  • Provides trend analysis for SLO performance over time
  • Creates executive summaries for stakeholders
  • Delivers data-driven recommendations for improving service reliability
  • Supports all entity types such as applications, websites, synthetic tests, and infrastructure

For more information, see AI-powered SLO analysis.

GenAI evaluation provides a comprehensive framework for assessing and monitoring the quality, accuracy, and reliability of LLM outputs in your AI-powered applications.

What GenAI evaluation does:

  • Enables systematic quality assessment of LLM model responses
  • Provides pre-built evaluators such as accuracy, relevance, completeness, logical consistency, and clarity
  • Supports custom evaluator creation for specialized use cases
  • Monitors pass and fail rates, scores, and detailed evaluation outcomes
  • Integrates with trace collection for real production data evaluation

For more information, see GenAI evaluation.

Getting started

To use Instana's AI capabilities, follow the setup steps for your environment type:

SaaS environments
AI capabilities are available by default with pre-configured LLM gateways. You can optionally configure custom gateways for your own runtime.
Self-hosted environments
  1. Enable the required feature flags for each capability.
  2. Configure LLM gateways to connect to watsonx.ai or vLLM-compatible services.
  3. Set appropriate user permissions.

Step 1: Enable the required feature flags (self-hosted environments)

To access AI capabilities on self-hosted Instana environments, enable the feature flag for each capability that you want to use:

Step 2: Configure supported LLM models (self-hosted environments)

Instana's AI capabilities are powered by the following large language models:

  • ibm-granite/granite-3.3-8b-instruct
  • mistral-medium-2505
  • mistralai/Mistral-Large-Instruct-2407
  • openai/gpt-oss-120b

Different capabilities can use different models optimized for their specific use cases. Configure an LLM gateway to the services that you want to use for Instana's AI capabilities. You can configure model selection through the AI gateway.

Step 3: Set up user permissions (all environments)

Access to AI capabilities is controlled through user permissions. Set the required permissions for users to access the AI capabilities that you want to use:

Access all gen AI capabilities
Enables access to AI-powered features on the Instana UI.
Access AI gateway
Enables read-only access to the AI gateway UI.
Create, configure, and delete LLM gateways
Enables full access to AI gateway management.

For more information, see Managing user access.

Troubleshooting

If you see an error message while you are using an AI capability, such as LLM gateway not found, LLM gateway disabled, or a similar gateway-related error, verify that the configured gateway is valid and enabled on the AI gateway page.

In addition, verify the following settings:

  • The gateway is correctly configured for the AI capability that you want to use.
  • The required feature flags are enabled for self-hosted environments.
  • You have the required permissions to access the AI capability and AI gateway.