IndiaFinance and insurance2026 – ongoingDezerv
Speaker-labeled transcripts for client conversations
Sarvam reported around 20% lower transcription word error rates at Dezerv
Problem
Dezerv's outsourced transcription pipeline struggled with multilingual investment conversations. The firm could not readily audit quality or search across its client calls.
Approach
Dezerv integrated Sarvam's speech-to-text SDK for client meetings and phone calls. With client consent, recordings passed through a batch API that separated speakers and returned structured transcripts. Relationship managers could search previous discussions without replaying recordings.
Results
Takeaway
Control over the transcription pipeline gave Dezerv direct access to quality measurements and searchable client histories. The reported improvement concerned transcript accuracy, not investment performance.
Categories
- Industry
- Finance and insurance
- Business function
- Customer serviceInternal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Conversational and voice AI
- Adoption
- Vendor solution
- Evidence
- Vendor case study
Sources
Source summaries
Sarvam described Dezerv's production transcription workflow and speaker separation. It reported around 20% lower word error rates than the previous setup.
Full details are in the original sources.