Bajaj Finance disclosed in its [Q4 FY26 earnings call] and [investor presentation] that it plans to deploy 600-plus autonomous agents in FY27, a quiet reduction from the 800-plus agents it guided for in Q3. The company gave no explanation for the revision.
Separately, it disclosed for the first time that it has completed a governance framework aligned with RBI’s FREE-AI policy, the Reserve Bank of India’s non-binding recommendations for responsible AI use in financial services.
The company has branded its AI strategy as FinAI and calls itself “BFL 3.0, A FinAI Company,” framing AI as central to long-term vision, not just an operational add-on. The enterprise-wide data architecture to support this strategy was completed in Q4 FY26. “We are not planning. We are in the process of implementing,” vice chairman and managing director Rajeev Jain said on the call.
Here are the key AI developments:
1. 800 to 600: autonomous agent target cut, sales scope dropped
In Q3 FY26, Jain said the company would have “800+ autonomous agents across sales, operations, HR, IT, risk and Debt Management Services (DMS) in the next fiscal.” In Q4, that number came down. “In FY27, we expect to deploy 600-plus autonomous agents and across operations, DMS, HR, technology and risk,” Jain said, with sales notably dropped from the scope. The investor deck shows 27 autonomous agents are currently live, with 118 agentic AI use cases planned for FY27.
Jain was categorical about what Bajaj Finance is attempting. “There’s a principal difference between deploying use cases and transformation,” he said. “Fundamentally, transformation is about reshaping the business model and, in general, covers all aspects of business. I just want to highlight that we are doing a transformation, and we are not just doing use case deployment.”
2. FREE-AI governance framework completed, but policy remains non-binding
“You’re aware that RBI has released a FREE-AI framework. So we have consumed, and we are in the process of implementing the explainability, auditability and transparency that the FREE-AI report talks about. And parallelly, a lot of investments are going into the security architecture as well,” said Anurag Chottani, Head of Corporate Strategy at Bajaj Finance, who leads AI implementation at the company. The investor deck states the company completed a governance framework aligned with FREE-AI in Q4 FY26 and plans to deploy governance policy across eight AI domains in FY27.
FREE-AI, released by RBI in August 2025, contains 26 recommendations across six pillars: infrastructure, policy, capacity, governance, protection, and assurance. It is not yet binding, with several provisions intended to be incorporated into Master Directions to make compliance mandatory.
Bajaj Finance’s commitment to a non-binding framework while deploying 600 autonomous agents making lending decisions raises the question of whether FREE-AI’s recommendations are adequate for the scale of deployment the company is describing. The deck also flags that an AI-driven fraud detection framework has been defined but not yet deployed, marked only as “initiated” for FY27.
3. One-third the cost: AI agents replacing human call centre staff
“AI call center agent is one-third of the cost, straight,” Jain said. The company had 5,000 outbound voice agents at peak and has actively brought that number down, and 30% of the remaining agents are now AI voice agents. As per the investor deck, 10 outbound and inbound AI voice bots and 17 AI text bots are currently live, with 72% of customer self-service now handled through AI voice and text bots. AI bots generated Rs 1,895 crore in disbursals in Q4, with the company targeting Rs 12,103 crore in FY27.
4. Debt collection via AI bots: from zero to 6,632 receipts in one quarter
“By June ’26, any communication across any channel by us as a firm, whether for sales, service, DMS, which handles loan collections — would have a text AI BOT in it,” Jain said. “So across 32 different lines of businesses that we run, plus a whole host of services. And we send, as I’ve said in the past, 260-odd million pieces of communication on a monthly basis. All of them will have a bot embedded in it.” The investor deck shows 6,632 debt management receipts collected via AI text bots in Q4, up from zero in Q3. Promotional text messages with AI bot penetration stood at 11% in Q4, targeting 100% in FY27.
5. 6x jump in Diwali loan processing, targeting 1 million per day next year
“Last year, we processed – on the peak of Diwali, on a single day, we processed like 600,000 loans, which, without AI, we were like 100,000 loans — we had capacity to process only 100,000 loans on a single day,” Chottani said. “So you might see next Diwali, like we’ll process close to 1 million loan accounts on a single day.”
6. 31 million voice interactions converted to data for loan targeting
Per the investor deck, 31 million voice interactions and 0.5 million text interactions were converted to data in Q4, generating 0.1 million new loan offers. The call gave no details on whether customers are informed that their voice interactions are being converted into data used for credit or marketing decisions.
7. Two separate camera deployments, no consent framework disclosed for either
The investor deck shows 60 face recognition cameras deployed in stores in Q4, targeting 1,000 in FY27, with 6.3 million existing customer face matches completed in Q4 and a target of 27 million in FY27. Separately, Jain described a broader rollout on the call. “Walking into 3,000 stores or 2,700 across branches – gold loan branches, customer service branches and in-stores – we will deploy close to 2,700 cameras, customer identification, and reduce friction, which means faster turnaround, which means lower cost,” he said. Neither the call nor the deck disclosed consent frameworks or data retention policies for either deployment.
8. Internal automation at high levels, per investor deck
9. 71,000 employees off PowerPoints, onto Copilot training by the end of May
“We have 71,000 employees. So far, we used to train using PPTs. I spent 33 years working. They’ll all transition by the end of May to Copilots. They’ll all use Copilots, which is an NLP-based infrastructure to get trained,” Jain said. On marketing, he said, “Almost 100% of our videos are now AI-generated.”
10. Customer discovery via ChatGPT and Gemini planned for FY27
The investor deck discloses the following pipeline for FY27:
11. “No concept of AI talent”, says Jain; company retraining machine learning staff instead
“Principally, there’s no concept of AI talent. These are existing machine learning (ML) folks who are actually being trained. They’re training themselves, and we are helping them with use cases to convert into AI researchers or AI talent,” Jain said. The company has 203 dedicated AI staff, expanding to 363 by June 2027.
Key Operational Numbers:
Also read:
For You
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。