AI In Banking: From Tokens To Talkbots

Digitisation has changed the banking experience dramatically over the years, doing away with the need for visiting a branch for everyday transactions. But even digitisation has its limits. Enter AI, which is making banking not just convenient but smart. Chatbots answer questions instantly, AI checks for fraud and your eligibility for loan within seconds. What took hours at a branch earlier now happens in minutes with AI.

The use of AI in banking operations began around 2016, when ICICI Bank introduced “software robotics” across more than 200 business processes. These systems emulate human actions to automate repetitive, high-volume, and time-consuming tasks across multiple applications. In 2017, State Bank of India (SBI) introduced SBI Intelligent Assistant (SIA), an AI-based smart chat assistant, that resolves queries of non-resident Indian customers. In 2018, Bank of Baroda also introduced an AI-enabled robot, Baroda Brainy.

According to a study released by the Reserve Bank of India (RBI) in December 2024, based on the annual reports of 32 commercial banks for eight years, FY2015-16 to FY2022-23, discussions on the use of AI in Indian banking date back to 2015-2016.

However, banks are still in the early stages of AI adoption. “I believe the banking industry will inevitably move in this direction. In the next 5-10 years, most banks are likely to operate more like technology companies with banking licences,” says Bhavnish Lathia, chief technology officer, Kotak Mahindra Bank.

Let’s assess AI’s impact on banking.

Expanding Capabilities

According to an EY report, The Idea of India 2025, “Big banks have taken up large-scale implementation projects focused on GenAI in the last few months of 2024. These projects talk of using GenAI in areas from cyber security copilot and corporate lending underwriting copilots to AI driven customer care platform implementations and on-premise GPU cloud deployments.”

Banks are also augmenting their in-house capabilities of AI as part of risk management and to ensure safety of data. According to the SBI Annual Report 2024-25: “SBI has developed a diverse portfolio of models covering most functions of the bank, including end-to-end digital loans, marketing leads, risk monitoring, operational efficiency and fraud prevention. SBI is using the GenAI Paradigm wherein two GenAI chatbots are assisting staff in customer service and compliance. GenAI ‘Ask SBI’ chatbot and ‘Deceased Case chatbot’ are helping staff in resolving queries.”

Private banks are also developing their in-house capabilities. Says Lathia: “We have built our own in-house AI platform, hosting large language models (LLMs) locally so that customer data remains secure. This required hundreds of engineers and millions of lines of code, but it ensures privacy without sacrificing the benefits of AI.”

However, the big question is if banks with weaker balance sheets can afford AI models? The RBI study cited earlier highlighted that banks with stronger financial health are more enthusiastic and proactive in adopting AI. Says Lathia: “Not every bank can do this, since it needs top tech talent. We’ve been hiring from Google, Microsoft, Amazon, Goldman Sachs, Netflix, Uber, and building this capability over three years. Regulators are also watching closely to balance convenience with privacy and security. Today we run around 20 foundational models in-house with dozens of AI agents supporting employees.”

Why Are Banks Adopting AI?

The most obvious reason is cost efficiency and enhanced productivity. Says Lathia: “AI helps augment what humans are doing whether in branch processing or risk models. It allows people to focus on higher-judgement tasks, while AI handles routine, repetitive ones. This improves productivity in a big way.”

The EY report also highlights that adoption of GenAI has the potential to enhance productivity of banking operations by 44-46 per cent, sales and customer service by 38-40 per cent and credit and collections 34-36 per cent, which can together account for 40-42 per cent of productivity improvement.

Customer Convenience: Lathia says AI can make smarter recommendations—from rebalancing stock portfolios to suggesting the right credit card—helping banks connect more meaningfully with clients. “Customers today want experiences, not products being pushed at them,” he says.

In December 2024, IDFC FIRST Bank launched an AI-powered holographic digital avatar of actor Amitabh Bachchan for customer engagement. The Holographic Extended Reality (HXR) device has touch capability, which allows users to interact directly with the digital avatar and get information on bank products and services, monthly interest credits, mobile banking, and their current accounts. The bank has also developed an AI-based bank statement analyser that offers insights to customers on their spending habits, inflows, and transactions.

AI also helps in solving the language problem in a linguistically diverse country like India by operating in local languages.

Better Grievance Redressal: As the number of customers are growing, so are the number of complaints. Promptly addressing complaints not only resolves issues but also prevents reputational risks, builds trust, ensures customer satisfaction and helps avoid regulatory penalties.

AI has stepped in here, too. While addressing the annual conference of the RBI Ombudsmen in March 2025, RBI Governor Sanjay Malhotra said, “AI has the potential to revolutionise grievance redressal. We are entering an exciting era where technology, particularly AI, can drive remarkable improvements in speed, accuracy, and fairness of complaint resolution.”

Leveraging data analytics, sentiment analysis, and predictive models, AI can be used to analyse large volumes of data to detect spikes in issues, such as ATM failures or erroneous charges, and alert regulated entities pre-emptively, he added.

Solving Fraud Detection: One of the major issues both banks and customers face is the issue of financial fraud. According to the National Crime Records Bureau (NCRB) data, online financial frauds constitute 67.80 per cent of all cybercrime complaints in the second quarter of 2022.

Mule accounts played a large role in this. It is a bank account used by criminals to move illegal money, either with the accountholder’s collusion or by tricking them with promises of easy cash. Earlier, banks used a rule-based system to detect mule accounts, but it look a long time to review them. That problem was solved when the Reserve Bank Innovation Hub (RBIH), a subsidiary of RBI, built its own AI/ML-based system to spot mule accounts. The new system uses machine learning to study transaction patterns and account details, and can predict suspicious accounts faster and with more accuracy.

Checks In Place

While AI presents unparalleled opportunities, it entails some risks as well. Successive RBI governors, including the incumbent one, have flagged this.

Malhotra said at a function in March 2025: “We need to be cognisant of the challenges and risks that its (AI’s) adoption poses. We must also remain mindful of ethical considerations. Human oversight, bias mitigation and data privacy must be integrated into the AI systems to ensure transparent and consistent outcomes.”

In order to tackle the concerns related to AI adoption, RBI constituted a committee in December 2024 to frame guidelines for the responsible deployment of AI in the financial sector; its report was released on August 13, 2025.

The report defines seven guiding principles, referred to as the Seven Sutras, which are meant to form the basis of the application of AI in the financial system. These are: trust is the building block, people first, innovation over restraint, fairness and equity, accountability, understandable by design, and safety, resilience and sustainability.

The committee has provided 26 implementable recommendations organised under six strategic pillars: infrastructure, policy, capacity, governance, protection and assurance. The committee has proposed that RBI should adopt a supervisory model striking a balance between innovation and risk management. It proposes intensive monitoring of AI systems.

Any change is accompanied by concerns. The same is happening with AI adoption, but with the correct measures in place, things are likely to fall into place.

AI And Its Uses

Process Automation

Faster KYC, compliance and paperwork handling.

Personalisation

Tailored financial advice and product recommendations.

Customer Service

Chatbots and voice assistants for 24x7 support.

Fraud Detection

Spotting suspicious patterns in real time.

Risk Management

Smarter credit scoring and loan approvals

kundan@outlookindia.com