Orca Security launches Discovery AI for cloud analysis

Orca Security launches Discovery AI, a tool that turns questions into filters for cloud analysis. The solution uses natural language and AWS models to simplify complex queries.

Discovery AI converts questions into Sonar filters, using a nine-step pipeline with models such as Anthropic Claude Haiku 4.5, Amazon Nova 2 Lite, Amazon Titan and Cohere Rerank 3.5. The architecture cut costs by 90%, tokens by 93% and latency by 75% compared with an initial approach. The tool supports more than 900 cloud infrastructure models.

The solution uses AgenticRetrieveStream to search for information in natural language and combines documents to provide grounded answers. The tool offers improvements in coding, cybersecurity capabilities and integration with OpenAI APIs.

Lower costs and latency transform cloud analysis

Lower costs and latency transform cloud analysis: Discovery AI cuts costs by 90% by optimizing cloud queries. This means teams save time and resources by avoiding manual processes. The tool also cuts tokens by 93%, which simplifies search and avoids data overload. Fewer tokens also means lower processing costs and greater speed. The 75% reduction in latency makes responses faster, facilitating real-time decisions. These advances allow professionals without technical knowledge of query languages to work efficiently.

This ensures that queries are converted into precise filters. The partnership with AWS and Z.ai strengthens integration with cloud ecosystems. The tool is a practical alternative. Discovery AI is a solution that uses natural language to query cloud models, without requiring technical knowledge. Discovery AI replaces technical queries with natural-language questions, reducing costs and latency in cloud analysis processes. The tool converts questions into Sonar filters, allowing users without technical knowledge to explore cloud infrastructure data. AWS presents an example of how to use Amazon Bedrock Knowledge Bases to build a claims assistant that answers questions in natural language with citations.