Software & Platforms|Product Engineering|Enable Innovation

RoojoomAutomated call handling

A sequential pipeline caused 7-second delays. A native speech-to-speech model cut latency to 800ms, enabling fluid, human-like dialogue.

Jan 25, 2026|11 days ago

Key results

Avg Response Time
800 ms
Max Response Time
2 seconds

The company

AI platform for automated customer journey orchestration and voice assistance.

IndustrySoftware & Platforms
LocationTel Aviv, Israel
Employees11-50
Founded2013

Result highlights

  • Average voice response time cut from 4.5s to 800ms
  • Max voice latency reduced from 7s to 2s

The story

A software provider delivering customer engagement solutions to small and midsize businesses across telecommunications, insurance, and finance industries.

The legacy voice system relied on a sequential speech-to-text-to-speech pipeline that created significant latency, resulting in stilted conversations and customer interruptions. Synchronization issues often caused the system to mispronounce words or disconnect calls entirely.

Roojoom transitioned to a native speech-to-speech architecture using Amazon Nova Sonic, combining comprehension and generation into a single model to reduce processing steps. The team prototyped the solution in under two weeks using Amazon Bedrock for secure model access and data control. Complementary models handle post-call workflows by summarizing transcripts and tagging actionable items like meeting requests.

Quotes

Explore similar

Find AI opportunities for your
business context

Understand what's working with 2,275 recent AI case studies across industries. We structure things so you can find high-impact strategies for your exact context.

Graphic placeholder

Industries covered