Druva transforms data security using Amazon Bedrock AgentCore
Druva built DruAI, an agentic AI solution to help customers complete cyber investigations and recovery workflows. The solution uses a multi-agent system on AWS with AgentCore Memory, AgentCore Code Interpreter, Amazon Bedrock access to Claude models, voice capabilities with Amazon Nova Sonic, and Bedrock Guardrails. In production, DruAI coordinates 8 to 10 specialized agents across investigation, help, and recovery workflows.
- Organization
- Druva
- Industry
- Tech & Comms
- Location
- United States
- Published
- May 2026
Reported outcomes
68%
quantified impactAutomation & deflection
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1Agentic AI
- 2Customer Support Automation
- 3Cybersecurity
- Customers faced a manual, high-pressure process to interpret telemetry and documentation during security incidents.
- Earlier AI workflows still required people to piece together insights and decide next steps.
- Existing build, train, and deploy AI approaches required constant infrastructure management.
- Druva designed DruAI as a collection of specialized agents that coordinate cyber investigation, remediation, and recovery workflows.
- AgentCore provides secure serverless hosting, request routing to internal tools, session isolation, automated scaling, code interpretation, identity controls, and memory.
- Druva uses Amazon Nova Sonic for conversational voice capability and Claude models in Amazon Bedrock for different agent roles, with Bedrock Guardrails for protections.
- 68% of customer issues are resolved through agent workflows without human intervention.
- Cyber investigations that previously took 30 to 60 days are completed in minutes.
- Support resolution is 58% faster when human teams assist.
- There have been over 17,500 AI conversations across more than 3,000 users.
Architecture
DruAI is a multi-agent system built on Amazon Bedrock AgentCore. It uses secure serverless hosting with session isolation and auto-scaling, request routing, AgentCore Code Interpreter for analysis tasks, AgentCore Identity for scoped access to backend systems, AgentCore Memory for personalization and learning across interactions, Amazon Bedrock for model access, Amazon Nova Sonic for voice, Claude Sonnet and Claude Haiku for different agent roles, and Amazon Bedrock Guardrails for prompt-injection and malicious-input safeguards. Druva initially used LangChain and later transitioned to Strands Agents SDK within the AWS environment.
Implementation partners1
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?