Real outcomes, on real books.
Every case study on this page carries a named client, a live production system and an attributable outcome number. No mockups, no synthetic pilots — just what the platform moved, how it was measured, and the engineering that carried it.
What the platform moved, where the client cleared it.
Reversing port-outs at scale with a Hinglish retention voicebot
Verified in-call retention from ~9% to a 48% peak on 4,700+ subscribers/day. Attributed only to confirmed CANCEL-to-1900.
Read the case study →Scaling insurance renewals with Asmita, a bilingual GenAI voicebot
250,000+ automated attempts. Live IPDS function-calling for IDV changes on the call itself. 24/7 outreach in the T-45 to T window.
Read the case study →AI voicebot collections program drives measurable recoveries
5,171 of 21,796 accounts recovered in month two. Real-time payment gateway completes collections inside the call.
Read the case study →Powering one partner's multi-brand collections voice channel
70+ bot builds across 16+ NBFC and bank brands, 300 concurrent calls, monthly volume up ~17× since ramp-up.
Read the case study →One platform, three journeys: consumer, commercial, dealer
Self-serve support +135%, 96–97% intent recognition held through a 43× engagement spike.
Read the case study →Four more case studies publishing this quarter
Loan servicing, MNP retention, diagnostics support, and consumer-durables collections — publishing through Q3 2026.
Notify me →Every number here is tied to an event we can point at.
Retention is counted only on the CANCEL-to-1900 SMS being confirmed on the call, not a verbal "yes". So the number understates the true business impact, deliberately.
Renewals are counted only when the payment settles, not when the customer accepts on the call. So the number tracks money in, not intent.
Collections are counted only against paid/unpaid dispositions per bucket per language, with the paid figure fed by the payment gateway in the call itself.
Multi-audience outcomes are attributed by bot, channel and month, so category-level gaps surface rather than blend into one flattering headline.