Godrej Capital becomes ‘Saksham’ with GenAI

Starting with business problems

At Godrej Capital, our Generative AI (Saksham) journey began with real business challenges. We identified three categories where we felt we could create meaningful impact.

First were the core business problems, like the heavy manual effort in document analysis, onboarding workflows, and underwriting, areas full of untapped automation potential. Second, we wanted to enhance individual productivity. Whether you're in sales or operations, there are routine tasks like reading documents or responding to emails that eat up valuable time. Third, we looked at process transformation. Instead of applying GenAI to individual pain points, we explored how the entire journey from start to finish can be reimagined?

Our mindset while devising Saksham, Godrej Capital’s in-house GenAI platform was, that it shouldn't just be a tool, but a business enabler.

It's not just GenAI 

A key learning very early on was that GenAI cannot be applied in isolation. You can’t drop a large language model into a broken process and expect transformation. What’s needed is a holistic view. One that combines process redesign, traditional AI/ML models, and GenAI to truly modernize how a task is done.

For instance, we realized that before using GenAI to summarize documents, we first needed to fix the upstream document handling process and plug in ML models to extract structured data. GenAI was the final enhancement, not the first solution. This combined, layered approach is what made the outcomes meaningful.

Today, the industry is evolving toward Agentic AI, autonomous agents that go beyond scoring or summarizing to actually performing actions. At Godrej Capital, we’ve begun moving in that direction, with GenAI tools that assist humans, while still keeping a human-in-the-loop for reliability and trust.

Saksham handles it all

Without control and standardization, GenAI can quickly become chaotic. Every department may want to build its own chatbot or experiment with its own models and before you know it, you're looking at security risks, rework, and fragmented adoption.

That’s where Saksham has made all the difference. It’s not just a platform for building GenAI use cases but a centralized foundation that ensures security, consistency, and reusability across the entire organization. Whether it’s sales, underwriting, or HR, every team can build on the same platform, choosing the right LLM for their use case, with data privacy, logging, and access controls already built in. The way query resolution has been structured is every department in Godrej Capital, similar to any GenAI platform, has a front end to fire queries. It then goes to Saksham, which does the processing and then returns the resolution to the front end.  

This has helped us avoid duplication, maintain governance, and accelerate innovation. Teams no longer need to worry about LLM connectors or cost thresholds. Saksham handles it all behind the scenes.

Saksham’s early successes

Let me share two specific examples. First, our contact centre operations. Earlier, analysing tens of thousands of customer calls for quality checks took months. With Saksham, we now use voice-to-text conversion combined with GenAI to evaluate calls in real time and generate automated feedback for agents. This feedback loop, which was once manual, is now embedded into daily operations.

Second, our multilingual training content. We needed to repurpose learning material in regional languages but doing that traditionally was expensive. Through Saksham, we now generate localized content at one-tenth the cost, in record time, while maintaining contextual accuracy. And all of this is done with one or zero humans actively reviewing, which shows how mature and stable the technology has become under controlled conditions.

Agentic AI, next frontier

Looking ahead, our big focus is Agentic AI, particularly in customer onboarding and credit underwriting. These journeys often involve too much back-and-forth between departments, slowing down the customer experience.

We're currently running proof-of-concepts (PoCs) with intelligent agents that can read documents, anticipate next questions, and pre-validate inputs. All while keeping a human in the loop for oversight. These agents are being developed within Saksham, and our goal is to move to production by October–November 2025. Early results are promising, and we believe this will take our GenAI journey to the next level.

Watch what’s under the hood

If there’s one learning I’d share with other CIOs, it’s this: don’t stop at development, focus on post-launch governance.Monitoring things like token usage, inference costs, and LLM performance drift is critical. The space moves so quickly that what worked last month may not be cost-effective tomorrow. Build in alerting and configurability so that you can react in real time.Also, be prepared for regulatory shifts. With laws like India's Digital Personal Data Protection Act, 2023 (DPDP Act), you must have data privacy safeguards embedded at the platform level. Your GenAI platform should never expose sensitive data unnecessarily and for that, a strong platform engineering team is non-negotiable.