How Kotak Mahindra Bank is Building An Operating System For The Future

Kotak Mahindra Bank continues to make heavy investments in technology,  “We continue to spend on technology about 12% to 14% of the total Opex that continues,” said the bank speaking with analysts after the Q2, 2025-26 results. The focus on Artificial Intelligence (AI) has also increased, with the bank working on a proprietary platform that will provide a cognitive base to the entire digital ecosystem. Besides the front end platforms being powered with AI, it is also being embedded for the backend tools targeted for the developers.

The bank also requires a strong backend for data processing for which a Data Exchange (DEX) platform has been built. 

Kotak AI Platform

The Bank is building Kotak AI, a proprietary Generative AI platform that will serve as the cognitive core of the Bank’s ecosystem. It will provide predictive insights, contextual intelligence and seamless automation across interactions, according to the bank’s annual report for 2024-25. 

The Bank is working on 'Koder' - An AI-driven development tool for software engineers and a conversational AI assistant for frontline and sales staff viz. 'Kompanion', says the report.

Data EXchange (DEX) Platform

Kotak’s digital transformation will be powered by a unified data platform for smart banking, the cloud-native DEX platform. It will serve as the intelligence backbone purpose-built on modern cloud architecture.

DEX will power everything: Hyper-personalised experiences to intelligent risk decisioning. DEX has unified structured and unstructured data across business lines into a single, trusted platform, as mentioned in the annual filing.

FE CIO speaks with Bhavnish Lathia, CTO, Kotak Mahindra Bank on how the bank is working on integrating AI at the bank and the plans for H2, 2025-26.

Edited Excerpts

Considering the importance of AI and the kind of usage that it's going to have across the fabric of how banking is done, what’s in the works at Kotak Mahindra Bank on the AI front ? 

Artificial Intelligence is a huge focus area for us. Both in terms of revenue efficiency, which is being able to drive more sales, cost efficiency, as well as customer experience. 

When a bank looks at AI, there are a few important challenges. The first is, we cannot allow our data to cross our firewall boundaries considering the factor of customer trust. 

Secondly, the LLMs in the market are trained on trillions of tokens but not specialized for the needs of the financial services or banking industry. So those models tend to hallucinate. Obviously, we want to make sure that the LLMs are accurate and not hallucinating.

For these two reasons, we embarked on a journey to build our own in-house AI platform. And at this stage, we run about 12 foundational models in-house. And when these models are queried, we don't have to cross our firewall boundaries. Our teams have carefully curated and fine-tuned these models. We've distilled them down to what we care about from a financial services standpoint. We've made sure to train and anchor them on our data, building them on certain ground truths to prevent hallucination of these models.

Elaborate on the skillset library and the agentic AI components ?

A skill set library has also been built. Currently it has a repository of about 26, 27 skill sets. For e.g. Image generation as a skill set. It's a very specific set of skills we've built on top of our in-house AI platform. 

We then built our own AI agent Software Development Kit (SDK), which allows us to create our own AI agents in-house. Followed by the indigenously developed workflow platform, that allows these agents to collaborate with each other. And then the final layer on top isa governance layer. That allows us to closely monitor the activities of these agents. And ensure, they stay within the guardrails. Thus enabling us to demonstrate both compliance as well as explainability of the decisions that these agents are taking.

This is our AI stack. We built it completely in-house, using our engineering talent. Effectively, we're building the operating system for a bank of the future. So ultimately, over the next few years we expect to see more and more of these agents in production and expect them to collaborate with each other. Moreover, they also assist employees in taking over some mundane tasks, under supervision of employees. But hopefully this allows us to provide much better CX as well as drives cost efficiencies. 

What are Kotak Mahindra Bank’s information technology plans for H2?

We are going to continue to build on the initiatives undertaken in H1. There are a lot of work streams underway. Whether it's related to improving our customer experience, reducing turnaround time on service requests and making progress towards a true paperless branch. We will continue to digitise our processes, also upgrading the apps, whether it's 811 or Solitaire or our SME propositions.

However beyond H2, a huge opportunity we see, is leveraging the power of Kotak as a conglomerate and tying it all together digitally. So for a customer, there's a seamless digital experience whether they want to trade in stocks or do banking or buy an insurance policy. It's about bringing the power of the conglomerate together in a digital way and that will play out over the next couple of years.