Mahindra Finance has crossed 1 crore calls using AI-powered voice agents in 12 Indian languages, marking one of the largest reported deployments of multilingual voice AI in India’s financial-services sector.
The company has expanded its partnership with Bengaluru-based Sarvam AI, using the technology across customer sales, loan collections and employee engagement. Mahindra Group’s AI division developed the voice agents on Sarvam’s platform.
The scale is particularly relevant for Mahindra Finance because much of its business extends beyond large metropolitan centres.
The lender operates through more than 1,348 offices and had crossed 1.2 crore customers by March 2026. Its assets under management stood at ₹1.34 lakh crore at the end of that financial year.
Proposal for Conducting Cyber Crisis Drill, Tabletop Exercise (TTEx) & CCMP Readiness Exercise
AI callers are moving beyond simple customer-service menus
Traditional automated calls usually follow a fixed script.
A customer hears a recorded message, presses a number and moves through a predetermined menu. The system cannot easily understand a natural response such as, “I cannot pay today, please call me next Monday.”
Voice AI is designed to make that interaction more conversational.
Mahindra Finance says its Sarvam-based agents can speak with customers in their preferred language and are being deployed for sales as well as collections. That means AI can handle conversations that previously required large teams of human callers.
Collections are especially important for lenders.
A finance company may need to contact lakhs of borrowers about upcoming or missed EMI payments. Automating part of that process allows it to make far more calls without expanding a call centre at the same rate.
Mahindra Finance had already disclosed before the latest announcement that AI voice agents were covering around 15% of its “soft bucket” collections portfolio. The company said AI-led collections had helped lower digital costs, reduce EMI bounce rates and expand coverage.
A soft bucket generally refers to borrowers in an early stage of payment delay, rather than accounts that have already fallen into serious or prolonged default.
Why 12 Indian languages matter for financial AI
India presents a particular problem for voice automation.
A system that performs well only in English may have limited value when a lender serves customers across small towns and rural markets where people are more comfortable discussing money in Hindi, Marathi, Tamil, Telugu or another regional language.
Sarvam has built much of its business around solving that problem.
The company says its voice technology is already being used at very large scale across India. Sarvam co-founder Pratyush Kumar said at GFF 2026 that the company handled 32.5 crore minutes of voice AI during the past year. Its infrastructure is currently processing around 40 crore API calls per day.
Earlier this year, Sarvam, EkStep Foundation and AI4Bharat also used multilingual voice agents to reach around 50 lakh people in 31 days across healthcare, agriculture and governance programmes.
That scale is helping establish voice as a major route for AI adoption in India, rather than relying entirely on text-based chatbots.
What are voice AI agents and “sovereign AI”?
A voice AI agent combines speech recognition, an AI model and speech generation.
It first turns what a caller says into information the system can understand. The AI decides how to respond, and a speech model converts that response back into a spoken voice.
Unlike a recorded IVR menu, an AI agent can potentially interpret different ways of asking the same question and continue a conversation.
Sovereign AI, meanwhile, refers broadly to AI infrastructure and models that a country or organisation can operate with greater control over where computing and data are located.
Sarvam positions itself as an Indian sovereign-AI company. It is building models, voice systems and computing infrastructure intended to operate at Indian scale and support local languages.
For banks and lenders, that pitch can become important because financial conversations may involve sensitive customer information.
Scale also creates questions about trust and disclosure
The opportunity is significant, but financial AI cannot be judged only by how many calls it completes.
Customers need to know when they are interacting with an automated system, particularly when the conversation concerns repayments, loans or other financial obligations.
Accuracy also becomes critical. A mistranslated repayment date or incorrectly understood customer response can create consequences that are very different from an error in an ordinary shopping chatbot.
The RBI has itself warned that growing AI use in financial services can amplify risks involving speed, concentration of technology providers and the opacity of automated systems. Deputy Governor Rohit Jain told GFF 2026 that automated systems can analyse information and initiate actions faster than people can respond.
There is also the problem of fraudulent calls impersonating lenders.
Mahindra Finance says it has transitioned legitimate service calls to the 1600 number series and promotional calls to the 1400 series, in line with telecom measures intended to make genuine financial-sector communications easier for customers to identify.
The Mahindra-Sarvam deployment therefore represents more than a call-centre efficiency project.
It is an early example of what financial AI could look like at Indian population scale: automated conversations, delivered cheaply, across several languages and integrated directly into everyday lending operations.
What this means for you: An AI voice calling about a loan or EMI should still be verified like any other financial call. Never disclose an OTP, PIN or banking password, and check that calls claiming to be from Mahindra Finance originate from its authorised 1600 or 1400 series.
The420 Insight: India’s strongest AI use case may not be another English chatbot. The bigger opportunity could be systems that speak naturally to millions of people in regional languages — but once those systems begin discussing loans and repayments, transparency, accuracy and customer protection become just as important as scale.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
