An NBFC’s collections calls in Marathi and Hindi
A voice agent handles early-bucket reminder calls, records promise-to-pay dates and hands hardship cases to a person. The success metric and its target are agreed in writing before the build.
AI for lenders, insurers and brokers who answer to a regulator.
We build collections agents, document readers and support agents for NBFCs, insurers, brokers, fintechs and co-operative banks. Every one is measured, logged and designed for the RBI, IRDAI and SEBI rules you work under.
The work that eats your teams’ time
Collections at volume
Thousands of early-bucket reminder calls a month, in Marathi, Hindi, Gujarati or Tamil. Agents read from a script, log the outcome by hand, and the promise-to-pay data ends up in a spreadsheet nobody trusts.
KYC and loan files
PAN, Aadhaar (masked), bank statements, ITRs, salary slips and property papers arrive as phone photos and scanned PDFs. Someone keys them into the LOS. CKYC checks and re-KYC reminders add another queue.
Claims documents
Insurers receive discharge summaries, bills, prescriptions and investigation reports in every format. Assessors spend most of their time finding information, not judging it.
Support that must be right
Customers ask about EMIs, foreclosure charges, policy status and statements on WhatsApp and email. A wrong answer is not just bad service. It can be a complaint to the ombudsman.
Legacy core systems
Many co-operative banks and older NBFCs run core banking or loan systems with no modern API. Any useful AI has to work with what is there: database views, batch files, screens.
What we build for financial services
Voice Agents in Indian LanguagesReminder and early-bucket collections calls in Hindi, Marathi and more, with disclosure at the start, a key press to reach a person, and every outcome written back to your system.See the service Document IntelligenceKYC, bank statement and claims documents read into structured fields, with a confidence score on each and a human check below the threshold you set.See the service Customer Support AgentsAnswers on WhatsApp, web and email from your approved policy content, with hand-off to a person for anything outside it.See the service Workflow Automation AgentsReconciliation, report preparation and checks that today run on spreadsheets and email.See the service AI Governance and DPDP ReadinessAI policy, risk assessments and controls that your board, auditors and regulator can read.See the service Managed AIWeekly quality measurement after go-live, so accuracy does not drift quietly.See the service
The rules we design for
This section is guidance, not legal advice. Your compliance team and counsel make the final call, and we work with them from the first week.
RBI
We follow the RBI’s outsourcing directions where we act as a service provider, and design for the principles in the RBI’s FREE-AI framework report (Framework for Responsible and Ethical Enablement of AI, August 2025): board-approved AI policy, human oversight, explainability and monitoring. Payment system data stays in India, as the RBI’s data localisation direction requires.
IRDAI
Insurers’ information and cyber security guidelines shape where data lives, who can access it and how incidents are reported. We deploy inside your environment and log every action.
SEBI
For brokers and intermediaries, we design around SEBI’s cyber security and record-keeping expectations, including retention of communications.
DPDP Act 2023
Customer data is personal data. We map every data flow, collect only what the use case needs, redact identifiers before they reach a model, and set retention limits.
What this means in practice: India-hosted deployments, AI disclosure at the start of every conversation, a human route always available, decisions on credit or claims left with your people, and a full audit log.
AI Governance and DPDP ReadinessRead: DPDP Act and AI agents
Illustrative scenarios
Read the scenario
A general insurer reading health-claim documents
Document Intelligence extracts the fields an assessor needs from mixed claim files, and flags missing documents on day one rather than day ten.
These are examples of what a project looks like. They are not clients or results.
Questions financial services teams ask
Will the AI make lending or claims decisions?
No. We build systems that prepare, check and summarise. Decisions on credit, claims and complaints stay with your authorised staff, and the system records who decided.
Where will our customer data be stored?
In your cloud account or on your premises, in Indian regions by default. We do not copy your data to our own systems, and it is never used to train models.
Can you work with our core banking or LOS vendor?
Yes. We prefer to work alongside your existing IT vendor. We integrate through whatever the system supports: APIs, database views, secure file transfer or, as a last resort, screen automation.
Is a voice agent allowed to make collections calls?
Automated calls must follow the RBI’s fair practices and recovery conduct expectations, telecom rules on calling, and DPDP consent. We build disclosure, call-time limits and an easy route to a person into the design, and your compliance team signs off the script.
How do you prove it works before we commit to a full build?
A fixed-price pilot of 4 to 8 weeks on your real data, scored against an evaluation set we agree in writing first. If it misses the target, you get the full report and are not obliged to continue.
Pick one queue. We will tell you if AI can clear it.
Thirty minutes with a founder. A one-page note within three working days.