North Carolina Division of Employment Security modernizes customer services with generative AI on AWS
The North Carolina Division of Employment Security (NCDES) implemented a generative AI-powered virtual assistant to modernize its unemployment insurance claim application process. NCDES aimed to improve speed and quality of customer support, reduce call center wait times, and provide 24/7 service access securely and responsibly. AWS services including Amazon Bedrock, Amazon Lex, AWS Lambda, Amazon OpenSearch Service, Amazon Comprehend, AWS Security Hub, AWS Key Management Service, and AWS Certificate Manager were used to build a secure, scalable, and compliant generative AI chatbot solution. The chatbot combines FAQ matching with foundation model-driven fallback responses to deliver accurate answers grounded in an internal knowledge base while protecting user privacy and minimizing hallucinations. Since launch in February 2025, the chatbot handled over 2,700 inquiries in the first month, reduced human intervention, improved citizen and employee experience, and provided insights for ongoing process and knowledge base improvements.
- Organization
- North Carolina Division of Employment Security
- Industry
- Public Sector
- Location
- United States
- Published
- May 2025
Reported outcomes
Strategic outcomes
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- North Carolina Division of Employment Security
- Provider
- AWS
- Maturity
- Unknown
- Linked source
- AWS Public Sector Blog
No explicit deployment-stage evidence found.
Primary read
Use case focus
Showing 2 of 2
- 1Generative AI Chatbot
- 2Customer Support Automation
- NCDES faced delays and limited service hours for unemployment insurance customer support due to traditional manual systems and aging infrastructure.
- The agency sought to reduce call center wait times, extend 24/7 availability, enhance the customer experience, and maintain data privacy and security.
- NCDES adopted a cloud-first strategy with AWS, building a generative AI chatbot in AWS GovCloud (US) to extend cloud modernization efforts.
- The chatbot uses Amazon Lex for conversational input understanding, Lambda for backend processing, Amazon OpenSearch Service to search FAQs, and Amazon Bedrock to generate fallback responses when FAQs do not match.
- Security and responsible AI principles were incorporated including PII masking, adherence to AI risk frameworks from NIST, and internal policies for trustworthy AI.
- The solution went live within four months, enabling consistent 24/7 customer self-service with a combination of AI-driven and curated knowledge base responses.
- The chatbot handled 2,700+ inquiries in the first month, reducing workload on human agents and call centers.
- Employees could focus on higher-skilled tasks while the chatbot improved knowledge management and customer experience.
- NCDES gained detailed insights from chatbot interactions to optimize both the online claims process and the knowledge base.
- The agency plans to expand AI-driven support to other public service areas, enhancing both citizen and employee experiences.
Architecture
The architecture uses Amazon Bedrock to access multiple AI foundation models, Amazon Lex for chatbot conversational interfaces, AWS Lambda for serverless function orchestration, Amazon OpenSearch Service for FAQ matching and vector-based knowledge retrieval, and Amazon Comprehend for language processing. Security is ensured through AWS Security Hub, AWS Key Management Service, and AWS Certificate Manager. The chatbot combines FAQ matching with generative AI fallback to generate accurate, privacy-preserving answers.
Sources & evidence1
- Customer explicitly identified
- Technical implementation details available
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?
Community
Comments
No published comments yet.