75% Automated. ₹2,744 Crore in AI-Linked Disbursements. Is Bajaj Finance Measuring the Right AI Success?
by 
Bajaj Finance is scaling AI across customer service, lending, and risk management. But one number reveals the difference between automating work and transforming the economics of a financial business.
What if the most important number in Bajaj Finance’s AI transformation is not the 75% of self-service volume handled by bots, but the financial metric that has barely moved alongside it?
In its Q2 FY27 earnings call on 19 October 2026, Bajaj Finance reported that AI voice and text bots handled 75% of do-it-yourself (DIY) customer-service volume, up from 65% a year earlier. Disbursements from leads generated by these bots reached ₹2,744 crore in the quarter, more than double the ₹1,295 crore reported in Q2 FY26.
Those are significant numbers. Yet consolidated operating expenses also rose 23% year on year to ₹5,275 crore, matching the 23% growth in net total income. Consequently, operating expenses as a percentage of net total income improved by only 10 basis points, from 33.1% to 33.0%.
That is not proof that AI has failed to reduce costs. It is a more interesting question: if automation is expanding rapidly, how much of its value is becoming visible in the economics of the wider business?
For leaders evaluating Conversational AI, this is the distinction between an impressive deployment statistic and a defensible business case. Bajaj Finance’s AI Bots provide an unusually useful case study because the company is reporting progress across customer servicing, commercial engagement, and agentic applications, not simply launching another chatbot.
Bajaj Finance’s AI Numbers Look Impressive. Three Metrics Reveal the Bigger Story.
The latest reported figures available as of 23 October 2026 reveal a broader transformation than the original 71% automation headline suggested.
Metric | Latest reported result | Why it matters |
DIY customer-service volume handled by AI bots | 75% in Q2 FY27 | Measures automation within the defined self-service category |
Comparable figure in Q2 FY26 | 65% | A 10-percentage-point year-on-year increase |
Disbursements from AI voice and text bot-generated leads | ₹2,744 crore | Connects AI-led engagement with lending activity |
Comparable AI bot-generated-lead disbursements in Q2 FY26 | ₹1,295 crore | Provides the baseline for year-on-year growth |
Consolidated operating expenses | ₹5,275 crore, up 23% | Shows how the wider cost base is evolving |
Operating expenses to net total income | 33.0%, versus 33.1% | Shows the change in the consolidated expense ratio |
AI agents and agentic applications | 129 agents and 65 applications, as reported on the call | Indicates that the programme extends beyond customer-service bots |
The data points to three separate outcomes: growing service automation, growing disbursements associated with AI-generated leads, and a consolidated expense ratio that has changed only marginally. These should not be collapsed into a single success score. Each measures a different part of the business, and each requires a different test of performance.
75% Automated Sounds Like a Breakthrough. What Exactly Is Bajaj Finance Counting?
First, what exactly is being automated?
Bajaj Finance’s reported figure refers to DIY customer-service volume. It does not establish that 75% of every customer contact, complaint, payment dispute, or interaction across the company is handled by AI. Nor does it establish that 75% of customer problems are resolved successfully without subsequent human intervention.
The denominator matters because routine service enquiries and complicated financial problems have different requirements. An automated answer about a standard process is not equivalent to resolving a disputed payment or handling a sensitive account-access issue.
This is where a Conversational AI strategy needs more than a headline automation rate. It needs a clear definition of eligible interactions, verified resolution, repeat contact, escalation, and customer effort.
Consider a simple example. A customer asks about a repayment date, receives an answer, and ends the conversation. That may be a successful self-service interaction. Another customer asks why a payment has not been credited, receives a generic explanation, and contacts support again the following day. If the first interaction is counted as contained, the automation metric may look healthy even though the customer journey remains incomplete.
For Bajaj Finance’s AI Bots, the 75% figure is evidence of the scale of self-service automation. To judge its service quality, readers would need additional outcome metrics that the headline figure alone does not provide. The distinction is critical for any lender planning to benchmark itself against the company.
AI-Generated Disbursements More Than Doubled. How Much Business Did AI Actually Create?
The more commercially interesting number may be the ₹2,744 crore in Q2 disbursements from leads generated by AI voice and text bots.
Compared with ₹1,295 crore in Q2 FY26, the latest figure represents approximately 112% year-on-year growth, calculated from the reported values. This is a substantial increase in lending activity associated with AI-generated leads. But disbursements are not the same as revenue, profit, or incremental value caused entirely by AI.
A lead may have converted through another channel. The cost of generating and qualifying that lead still matters. So do product suitability, credit quality, acquisition cost, and the repayment performance of the resulting loan portfolio. To demonstrate commercial value rigorously, a lender should track the complete funnel:
- Leads generated through AI voice and text interactions.
- Qualified leads that meet defined eligibility criteria.
- Applications submitted and approved.
- Loans actually disbursed.
- Incremental disbursements attributable to AI, compared with a credible baseline.
- Acquisition cost, risk-adjusted returns, and subsequent customer outcomes.
That is the difference between reporting AI-associated activity and demonstrating AI-attributable value. The same principle applies to Conversational AI in customer servicing. A high interaction count is an input. A verified resolution is an outcome. A commercial conversion is another outcome, and its financial contribution must be measured separately.
AI Is Scaling. Expenses Are Up 23%. Does Bajaj Finance’s Cost Data Tell a Different Story?
Bajaj Finance’s consolidated operating expenses increased from ₹4,297 crore in Q2 FY26 to ₹5,275 crore in Q2 FY27. Net total income grew from ₹12,982 crore to ₹15,974 crore over the same period. Both grew by approximately 23%, leaving the operating-expense ratio at 33.0%, compared with 33.1% a year earlier.
This creates the article’s central tension, but it must be interpreted correctly.
A nearly unchanged expense ratio does not isolate the impact of AI. Bajaj Finance is growing its business, adding customers, expanding technology capabilities, and investing in operations. Those activities affect the consolidated cost base, so the ratio cannot tell us how much more the company would have spent without automation.
Management also stated during the Q1 FY27 call that voice AI was operating at approximately one-third of human labour cost under its comparison framework. It explained that the company seeks to make AI workloads materially cheaper than the human labour they replace, while using techniques such as smaller models and internal retrieval infrastructure to manage costs.
That is an important unit-cost principle, but it is not equivalent to saying the entire service operation costs one-third as much. A full economic assessment must account for model usage, telephony, integration, monitoring, maintenance, failed interactions, repeat contacts, and human escalation. It should then compare the complete cost of resolving a customer issue before and after automation.
The best metric is cost per successful resolution, not cost per chatbot conversation. For Bajaj Finance’s AI Bots, the next useful disclosure would be a clearer bridge between the unit economics of automated servicing and the operating efficiency of the overall business. That would help investors and technology leaders distinguish early deployment benefits from savings that are visible at enterprise scale.
From 129 AI Agents to Credit-Risk Models: Bajaj Finance Is Building Beyond Chatbots
The wider AI programme makes this case more important than a customer-service story. In the Q2 FY27 earnings call, management reported 129 AI agents and 65 agentic applications, with plans to expand the deployment during FY27. It also described a customer-risk model being tested on 500,000 business-to-business accounts, with early modelling results indicating a 15% to 20% improvement in risk outcomes. Management said the results would need to be assessed through subsequent performance, including three-month-on-book outcomes, before wider rollout decisions.
These are different categories of AI. Customer-service bots interact with customers. Agentic applications can support or execute workflows. Risk models inform decisions about credit and potential losses. Each requires its own evidence and safeguards.
The reported risk-model improvement is a modelling result, not proof of a realised 15% to 20% reduction in credit losses. The distinction between model performance and actual business outcomes is just as important as the distinction between conversation containment and customer resolution.
A serious AI programme needs different scorecards for each layer:
AI application | Evidence that matters |
Customer-service bots | Verified resolution, repeat contacts, escalation quality, cost per resolution |
AI-generated leads | Incremental conversion, acquisition cost, disbursement quality, risk-adjusted returns |
Agentic workflows | Completion rate, exception rate, human intervention, auditability |
Risk models | Out-of-sample performance, early delinquency, credit losses, fairness and stability |
This is the more valuable lesson from Bajaj Finance’s AI transformation: the company is applying AI across multiple business processes, but success must be established separately for each application.
A Bot Can End a Conversation. Can It Actually Solve the Customer’s Problem?
A high automation rate is valuable only if the customer can complete the task reliably.
Imagine a borrower asks about an instalment and then explains that the payment has already been deducted from their bank account. A capable system should recognise that the interaction has changed from a routine enquiry to a possible payment-reconciliation problem. It should preserve the context, use approved information, and route the customer to the right process.
That is what a strong Conversational Experience requires: accurate understanding, useful action, appropriate boundaries, and a clear path to human help.
Lenders should monitor four outcomes particularly closely:
Resolution: Was the original issue actually completed, rather than merely acknowledged?
Repeat contact: Did the customer return about the same unresolved problem?
Escalation: Was human support provided when required, with the context preserved?
Customer effort: How many steps, transfers, or repeated explanations were necessary?
These metrics matter because automated interactions can appear efficient while moving work elsewhere in the organisation. A bot that closes a session prematurely may reduce its own reported workload but increase calls, complaints, and manual handling later. The objective is not to maximise automation at any cost. It is to automate suitable tasks, preserve human judgement where necessary, and improve the total customer journey.
The AI ROI Questions Every Financial Services Leader Should Be Asking at CACES Asia
The same questions extend across the financial-services industry. Organisations want to know how to scale AI, calculate returns, maintain customer trust, and avoid confusing impressive adoption figures with proven business results.
CACES Asia, the Conversational AI and Customer Experience Summit Asia, is relevant to that discussion because these challenges sit at the intersection of customer experience, enterprise technology, operations, and governance. The event’s official website lists 2027 editions in Kuala Lumpur on 15 April, Jakarta on 15 July, and Singapore on 20 August, with dates and venues subject to confirmation. For leaders attending CACES Asia, Bajaj Finance offers a useful set of questions to take into the room:
- How should companies define and disclose the denominator behind automation rates?
- What evidence proves that an AI interaction resolved the customer’s problem?
- How should AI-generated commercial activity be separated from genuinely incremental business?
- Which costs belong in the calculation of AI return on investment?
- How should organisations validate agentic workflows and AI-driven risk decisions?
These are more useful questions than asking how many bots an organisation has deployed. They help move the discussion from adoption to operational performance and measurable customer outcomes.
Conclusion: The Real AI Benchmark Isn’t Automation. It’s Proven Business Value.
Bajaj Finance’s reported 75% DIY automation rate and ₹2,744 crore in AI bot-generated-lead disbursements show that AI has become a meaningful part of its service and commercial operations. The wider rollout of agentic applications and risk models suggests that the programme extends beyond conversational interfaces.
But the consolidated expense ratio, which improved by just 10 basis points year on year, reminds us that a business-wide AI transformation cannot be judged from automation rates or disbursement figures alone.
The next proof points should include verified service resolutions, repeat-contact trends, fully loaded cost per resolution, incremental commercial contribution, and realised risk outcomes. These would help establish how much value the technology creates, where that value appears, and what customers gain from it.
For Indian lenders, the opportunity is not simply to replicate Bajaj Finance’s AI Bots. It is to build measurement systems that distinguish activity from outcomes and unit-cost improvements from enterprise-wide returns. That is the standard for Conversational AI in financial services: not how much work the technology claims to automate, but how reliably it improves the outcome and the economics of that work.
It is also the right question for CACES Asia: how can financial institutions turn AI deployment into demonstrable improvements in customer experience, operational efficiency, and responsible growth?
FAQ's
Bajaj Finance reported that its AI voice and text bots handled 75% of do-it-yourself (DIY) customer-service volume in Q2 FY27, compared with 65% in Q2 FY26. This indicates growing automation within its defined self-service category, not that AI handles 75% of every customer interaction across the company. To assess the impact, businesses must also examine successful resolution rates, repeat contacts, human escalations, and cost per resolution.
Bajaj Finance reported ₹2,744 crore in Q2 FY27 disbursements from leads generated by AI voice and text bots, compared with ₹1,295 crore in Q2 FY26. That is approximately 112% year-on-year growth based on the reported figures. However, these disbursements do not, by themselves, establish how much additional business AI created or how much profit it generated. Incremental conversions, acquisition costs, loan quality, and risk-adjusted returns are needed to establish the commercial contribution.
Conversational AI can reduce the cost of handling suitable customer interactions by automating repetitive tasks and reducing some manual workloads. During its Q1 FY27 earnings call, Bajaj Finance's management said voice AI operated at approximately one-third of human labour cost under its comparison framework. However, the total savings depend on model and telephony costs, technology maintenance, failed interactions, repeat contacts, and human escalation. The more meaningful benchmark is fully loaded cost per successful resolution, rather than cost per conversation.
Bajaj Finance reported an operating-expense-to-net-total-income ratio of 33.0% in Q2 FY27, compared with 33.1% in Q2 FY26. Both consolidated operating expenses and net total income grew by approximately 23% year on year. These figures do not isolate AI's contribution because the company's overall costs reflect business growth, technology investment, and other operational activities. A near-flat expense ratio neither proves nor disproves AI savings. Establishing the impact requires a comparison of the costs and outcomes with and without automation.
AI automation measures the share of eligible work handled by AI. AI-generated leads measure customer opportunities originating from AI-enabled interactions. AI-driven business value measures the incremental financial or operational improvement attributable to the technology. For example, a bot may automate a service enquiry without generating a lead, while an AI-generated lead may result in a loan that would otherwise have been acquired through another channel. Financial institutions should track service efficiency, incremental conversions, acquisition costs, and risk-adjusted returns separately.
Banks should evaluate Conversational AI using a combination of operational, customer, and financial metrics. Core measures include verified resolution rate, repeat-contact rate, escalation rate, fully loaded cost per resolution, incremental conversion rate, customer effort, and risk-adjusted financial contribution. Organisations should compare these measures against a credible baseline and account for implementation, integration, maintenance, and oversight costs. A high automation rate is evidence of adoption, but it is not sufficient proof of return on investment.
A successful Conversational Experience enables customers to complete their tasks accurately, securely, and with minimal effort. In lending, this means more than providing automated answers. An AI system should preserve conversation context, give appropriate information, recognise when a problem requires a different workflow, and escalate complex or sensitive cases to human support. Repeat contacts, verified resolution, escalation quality, and customer effort help determine whether automation improves the complete customer journey.
Conversational AI bots primarily interact with customers through voice or text, answering questions and supporting defined service or sales workflows. AI agents and agentic applications can go further by coordinating multiple steps, using approved tools, and carrying out parts of a workflow within defined permissions. In its Q2 FY27 earnings call, Bajaj Finance reported 129 AI agents and 65 agentic applications. Their business value should be assessed through task completion, exception rates, human intervention, auditability, and measurable operational outcomes.
Yes. AI can support credit-risk analysis by identifying patterns, prioritising reviews, and informing risk decisions. Bajaj Finance's Q2 FY27 earnings-call discussion included a customer-risk model being tested on 500,000 business-to-business accounts, with early modelling results indicating a 15% to 20% improvement in risk outcomes. Those early results should not be interpreted as a proven reduction in realised credit losses. Validation against subsequent repayment performance, delinquency, model stability, and appropriate governance is necessary before drawing that conclusion.
The key lesson is to measure AI outcomes across separate business functions rather than relying on one headline automation percentage. Customer-service bots require resolution and cost metrics; AI-generated leads require incremental conversion and risk-adjusted returns; agentic workflows require reliable execution and oversight; and credit-risk models require validation against actual portfolio performance. Bajaj Finance's reported results illustrate the scale AI can reach, while also showing why adoption, commercial activity, and enterprise-wide financial returns must be evaluated separately.
At CACES Asia, enterprise leaders can explore how organisations define automation rates, verify customer resolution, calculate the fully loaded cost of AI, attribute incremental revenue, and govern AI agents. These questions are especially relevant to banking and financial services, where customer trust, data protection, operational reliability, and credit risk are as important as efficiency. The central discussion should be how to turn Conversational AI adoption into measurable customer and business outcomes, rather than simply increasing the number of automated interactions.
Leaders evaluating enterprise AI, customer experience, and automation can explore CACES Asia, the Conversational AI and Customer Experience Summit Asia. Its programme and regional editions offer a relevant context for discussions about AI deployment, customer engagement, operational performance, and governance. Readers should consult the official CACES Asia website for the latest event dates, locations, and programme details.