Aivar partnered with the customer to design and implement an AI-powered IT service portal, leveraging AWS Bedrock Agents to automate their IT service management workflows. The solution replaced manual processes with intelligent, agentic automation and streamlined IT request validation and routing. By integrating advanced AI/ML capabilities, the platform ensured consistent, policy-driven outcomes, reduced human error, and freed technical teams to focus on strategic initiatives. The system was built to be extensible, supporting future integration with additional SaaS and AWS services as business needs evolve.

The architecture is anchored on AWS’s Serverless and AI services, ensuring high availability, scalability, and security:

• Jira serves as the user-facing ticketing and approval system, integrating seamlessly with existing ITSM workflows.
• Amazon API Gateway exposes secure RESTful endpoints for Jira webhooks, triggering event-driven processing.
• AWS Lambda orchestrates all business logic, including ticket validation, integration with external systems (GitHub Actions, Workato), and communication with Amazon Bedrock.
• Amazon Bedrock (Claude Sonnet 3.5) provides the GenAI layer for intelligent, context-aware ticket validation and decision-making, leveraging advanced prompt engineering and chain-of-thought reasoning.
• AWS Secrets Manager ensures secure, auditable storage and rotation of credentials for all integrated systems.
• Amazon CloudWatch delivers unified logging, monitoring, and alerting for all workflow components.  

The stateless, event-driven design enables unlimited parallel processing, with all actions logged for compliance and auditability. The architecture is resilient, with multi-AZ deployment, retry logic, and disaster recovery via Infrastructure as Code.

• Processing time for access requests dropped from 8 hours to 15 minutes—a 95% reduction—enabling near real-time response to business needs.  
• Operational costs related to access management were reduced by 40% through automation and serverless scaling.  
• Validation accuracy reached 98%, minimizing errors and rework, and ensuring compliance with evolving business policies.  
• 92% of access requests were processed end-to-end without manual intervention, with ongoing improvements targeting full automation.