Customers

Deployments, numbers and what changed.

Institution, volume, the regulation that shaped the design, and the result. IndusInd Bank and TeamLease among the names; detailed references under NDA.

Customers
IndusInd Bank
Banking · India
TeamLease
Workforce operations · India
Named references
Under NDA
5 countries
IN · UAE · US · VN · ZA
Deployments

Three segments. One way of reporting.

Private life insurer
Insurance · premium collections · Hindi + English

Bilingual agent for lapsed-premium and ECS-bounce calls.

Challenge

Reach every policyholder whose auto-debit bounced, within IRDAI norms — without the loops and repeated reminders of the previous bot.

Solution

Multi-stage agent, frozen disclosures in both languages, PII redaction, and a regression scoreboard before every release.

Result

add client-approved metrics Right-party contact, promise-to-pay, repeated reminders and containment, reported daily.

Languages
2
Hindi · English, parity-checked
Calls analysed
100%
per-call QA, daily report
Reminder loops
eliminated by stage design metric
Release gate
regression before every change
Public-sector bank · RFP
Banking · PSU · multilingual voice AI

Scoped for on-prem, priced for the tender.

Challenge

An RFP for multilingual agents inside the bank’s infrastructure, aligned with RBI IT-outsourcing directions and data localisation.

Solution

Self-hosted speech and language models on dedicated GPUs, a control mapping to RBI directions, PSU-ready commercials.

Status

update as the bid progresses Bid in progress.

Deployment
On-prem
dedicated GPUs, self-hosted
Data residency
India
inside the bank’s DC
Regulatory map
RBI
outsourcing · FPC · KYC MD
Languages
Multi
Hindi, English, regional
NBFC lender
NBFC · early-bucket collections · Hinglish

Early-bucket collections for an NBFC book.

Challenge

describe the client’s situation Collectors covered a fraction of the pre-due and 1–30 DPD book daily; scripted reminders were ignored.

Solution

Bucket-wise graphs with promise-to-pay capture, reason codes, WhatsApp payment links and RBI FPC calling-hour enforcement.

Result

add client-approved metrics Reported daily against a control group.

Book coverage
100%
callable daily, in window
PTP capture
metric
Roll-forward
vs. control metric
Cost per RPC
vs. dialer team metric
What every customer gets

The daily report, per-call evidence.

Every deployment reports the same way: the metric you chose, the baseline it is measured against, and the evidence for every call — stage path, frozen lines, flags.

How a pilot runs
Daily report · 2026-09-15 · collections-hi
Daily performance reportagent: collections-hi · 100% of calls analysed
Right-party contact
vs. dialer baseline metric
Promise-to-pay
right-party contacts metric
Repeated reminders
per 100 calls metric
Containment
resolved without a human metric
Per-call evidence
09:02call_0412hi-IN · verified → ptp_captured → close
09:05call_0413en-IN · already_paid → utr_lookup → close
09:07call_0414hi-IN · wrong_number → dnc_flag → close
09:11call_0415hi-IN · legal_threat → warm_transfer · flagged for review
09:12frozen_linesdisclosure · consent · grievance · exact-match on every call

Illustrative report layout. Client numbers replace the dashes after the pilot. add client-approved metrics

References in your segment, under NDA

Hear it from a peer.

Ask for a reference call
Pilot

See how a pilot runs

How a pilot runs