Intelsense Technology logo
All case studies

Case Study · A national telecom operator

Hearing 100% of subscriber calls, not a sample

Bangla-first speech AI transcribed and analysed every subscriber call, surfacing churn signals and lifting ARR 30%.

92%

Bangla speech accuracy

100%

Of calls analysed

+30%

Customer ARR

The challenge

The operator handled millions of subscriber conversations, but quality and churn teams could only manually review a tiny sample. The signals that predicted churn and dissatisfaction were buried in unanalysed audio.

Off-the-shelf speech tools handled English passably but stumbled on real Bangla and code-switched speech.

What we built

Intelsense's Speech AI — built on the SenseVoice engine — transcribed and analysed 100% of calls at 92% Bangla accuracy, tagging intent, sentiment and complaint themes automatically.

Quality and retention teams gained a live, searchable view of what subscribers were actually saying.

The results

With every call understood rather than sampled, the operator acted on churn and up-sell signals it had been blind to — growing customer ARR by 30%.