Case Study: Phonely achieves lower WER and higher qualification rates with ai-coustics

Phonely cuts WER from 24.9% to 22.7% with ai-coustics

Phonely, a voice AI company running structured lead qualification pipelines at scale, faced a significant challenge with audio quality. Their agents struggled with noisy call audio from real-world environments, which led to higher word error rates (WER), broken call flows, and lost revenue. To fix audio at the source without adding latency, they turned to ai-coustics for a solution.

ai-coustics implemented its Quail Voice Focus model to isolate the primary speaker and enhance speech. This resulted in a measurable drop in WER from 24.9% to 22.7% for an enterprise customer. More importantly, this improvement correlated with a 3.65% relative increase in successful qualifications and an 8.73% relative increase in billable outcomes, validating that better audio quality directly improved Phonely's production performance.

Phonely