AI agent retrieval: How Ada scales with Cohere
A case study of Cohere
- Client
- Ada
- Industry
- Customer Experience (ACX)
Client website: ada.cx
Ada used Cohere Embed and Model Vault to achieve 77ms retrieval latency and 90.7% recall for its agentic customer experience platform.
The challenge
Ada needed to reduce retrieval latency for its AI agents, particularly for voice-based customer service where unnatural pauses in conversation impact customer confidence.
The solution
Ada implemented Cohere Embed on Amazon SageMaker AI for online retrieval and used Cohere Model Vault for offline knowledge-base indexing and bulk ingestion.
The results
The implementation achieved a median retrieval latency of 77 milliseconds and a 90.7% top-10 recall, while reducing vector storage requirements by 4x without measurable degradation.
In numbers
- median retrieval latency
- 77 ms
- top-10 recall
- 90.7%
- greater throughput-per-dollar for offline embedding
- 2.7×
- reduction in vector storage footprint
- 4×
Products and services used
Added by AntSpace from public sources (cohere.com).