Portfolio
The solution for Plexus
Category:
Integrated technologies:
Amazon Bedrock (Claude 3, Sonnet, and Titan), AgentCore Framework (Runtime, Memory & Observability), Knowledge Bases for Amazon Bedrock (RAG), AWS Security Hub, Amazon GuardDuty, AWS Lambda, Amazon SES, Amazon CloudWatch
About the Company Plexus
Plexo faced two major operational challenges: in SecOps, the manual analysis of large volumes of security findings; and in Support, an overreliance on human experts for incident resolution.
The challenge
The security team was struggling with manually correlating findings from Security Hub and GuardDuty, which delayed the detection of patterns. Incident management relied on manual searches of historical emails and tickets, creating bottlenecks.
AWS applied in this challenge
Amazon Bedrock (Claude 3, Sonnet, and Titan): Language models for understanding natural language requests, decision-making, and generating intelligent responses.
Agentcore Framework (Runtime, Memory & Observability): The core component of the ecosystem for intelligent agent orchestration, session context memory persistence, and operational traceability and auditing.
Knowledge Bases for Amazon Bedrock (RAG): Historical Knowledge Base.
AWS Security Hub & Amazon GuardDuty: Security Data Sources.
AWS Lambda: Serverless functions (Lambda Messenger and Lambda Orchestrator) responsible for receiving system events and coordinating requests from external channels to the agent.
Amazon SES: Smart Report Distribution.
Amazon CloudWatch: Observability and Metrics.
The Technical Solution
To transform Plexo’s operations, a Proof of Concept (PoC) was implemented based on a cutting-edge serverless architecture that brought an intelligent agent to life using Agentcore Runtime. The solution automates key processes through two high-impact workflows:
- SecOps & Self-Managed Infrastructure: Amazon CloudWatch immediately detects risk events in the cloud infrastructure. The intelligent agent, powered by Amazon Bedrock’s LLMs, instantly analyzes the context and creates operational tickets in Jira for resolution based on the assigned priority.
- Real-Time Conversational Support: An advanced chatbot has been integrated into Slack, allowing users to interact with the agent using natural language. Thanks to Agentcore Memory, the agent retains the context of the conversation to instantly view, read, or create tickets in Jira, transforming the chat into an operational control console.
The results
90%
90% reduction in the time required to analyze security reports.
Automatic generation of reports with recommendations prioritized by impact.
100%
100% Contextualized
Intelligent response capabilities based on actual ticket and email history.
3-in-1
Centralized Operations: Slack, Jira, and Smart Memory
We connect key systems in one place: now teams can view, read, and create Jira tickets using natural language from within Slack. Plus, the agent remembers the conversation thread to provide seamless, context-aware responses.
