Modernizing Cyber Risk Intelligence: How Directeam Helped At-Bay Automate Security Assessment Analysis with AWS Generative AI
Meet At-Bay
At-Bay is the world’s first InsurSec provider designed from the ground up to help businesses tackle cyber risk head-on. By combining industry-leading insurance with world-class cybersecurity technology, At-Bay offers end-to-end prevention and protection for the digital age.
At-Bay is a globally distributed company with hubs in Atlanta, New York City, San Francisco, and Tel Aviv.
The Challenge: Manual analysis of security assessments limited scale and insight quality
At-Bay’s cyber risk evaluation pipeline originally relied on manual review and traditional OCR-based workflows to process uploaded security questionnaires and assessments. This approach created operational bottlenecks and introduced inconsistencies in how unstructured, multilingual content was interpreted. As both the volume and complexity of incoming documents grew, At-Bay needed a solution to automate and enrich document analysis while generating structured, high-quality risk insights.
The Solution: A Generative AI enrichment pipeline powered by Amazon Bedrock and Claude
To address these challenges, Directeam helped implement a scalable Generative AI-powered architecture centered on Amazon Bedrock, where Claude models serve as the intelligent enrichment layer.
The solution architecture includes:
- Amazon S3 for storing incoming customer-submitted documents.
- Amazon Textract to extract raw text from scanned assessments and questionnaires.
- AWS Lambda and Amazon SQS to orchestrate the extraction and enrichment workflows.
- Amazon OpenSearch and Amazon Aurora PostgreSQL to store enriched outputs, enabling downstream systems to consume both structured data and natural language insights in real-time.
- Amazon Bedrock with Claude models, used to:
- Enrich extracted text with contextual insights and human-readable summaries.
- Handle edge cases and perform validation by generating contextual responses for uncommon or complex data patterns.
- Run additional GenAI routines to support insight generation and multilingual document understanding.
By placing Claude on Amazon Bedrock at the center of the enrichment layer, the pipeline was transformed from a rule-based extractor into an intelligent GenAI agent capable of interpreting complex cybersecurity documentation with consistency and depth. This allowed At-Bay to automate insights that previously required manual review and deep domain expertise.
The Results: Scalable, accurate, and insight-rich cyber risk pipeline
- Eliminated manual extraction through full automation of risk document processing.
- Leveraged Claude to deliver accurate, context-aware cybersecurity summaries.
- Enabled support for multilingual content with no added complexity.
- Achieved higher consistency and reduced false positives in risk analysis.