Interview with Brij Pahwa, Editorial Lead, e4m and MarTechAI.com
GenAI is not just reshaping workflows inside technology teams. It is also changing what business leaders expect from CTOs, especially in industries like insurance where trust, speed, and empathy have to coexist.
In a conversation with Brij Pahwa, Editorial Lead at e4m and MarTechAI.com, Sriram Naganathan, President Technology and Operations and Chief Technology Officer at HDFC ERGO General Insurance, spoke about the evolving role of the CTO, why GenAI has accelerated that shift, and how the company is applying AI across customer experience, partner productivity, and claims operations.
From tech custodian to business co-creator
Naganathan said the shift in how organisations view technology leadership has been building for years, driven by the broader digital transformation wave, and GenAI has only accelerated it.
He linked the CTO’s expanded role to the centrality of data and insights in business strategy. In his view, technology leadership is no longer only about execution. It is also about understanding how AI can empower business outcomes, anticipating what is coming next, and helping architect the organisation for that future.
He also emphasised that the modern technology leadership role spans multiple priorities beyond core IT delivery, including innovation, cybersecurity, sustainability, and enterprise-wide transformation. For him, being close to the business strategy comes with a dual lens: managing risk while delivering value.
How GenAI is changing the CTO’s job
Pahwa asked how GenAI has shaped him as a CTO and how it has affected teams. Naganathan’s answer positioned GenAI as an acceleration layer for technology leaders. He described it as a “co-pilot” that helps CTOs connect signals, build intelligence from data, and apply learning from parallel industries.
He noted that, historically, technology approaches were often siloed by industry, but GenAI is pushing leaders to borrow ideas more freely across sectors. Insurance, he said, can learn from sectors like retail and supply chain, not only from allied financial services.
Where GenAI creates the most value in insurance
While operational efficiencies are already visible in areas like contact center operations and parts of marketing, Naganathan said the more meaningful value is created when GenAI starts improving core insurance moments and contributing to business outcomes.
He called claims the “moment of truth” for insurers and described it as the point where service is truly tested. He underlined that claims are also emotionally charged for customers, which makes experience design critical.
He discussed how GenAI can improve claims experience, assist servicing, and support fraud detection. He pointed to a practical customer pain point: companies often end up inconveniencing the majority of genuine customers due to a small fraction of fraudulent cases. In his view, GenAI’s promise is to narrow down that risky minority so the broader customer base faces fewer friction points.
Making policy language easier to understand
Naganathan shared an example from HDFC ERGO’s customer-facing ecosystem app, HERE, which he described as a super app launched about a year ago and now at more than a crore downloads.
He spoke about enabling the app with a feature called Know Your Policy, also referred to as KYP. The goal is to help customers understand what their policy covers through simple English prompts. He explained that the system scans the policy and supports questions that can be specific to a customer’s situation, such as how coverage applies for a pre-existing disease, when it becomes applicable, and whether deductibles are involved.
He described this as a trust-building move, because it reduces confusion and helps customers interpret complex policy wording more clearly.
Small language models, enterprise data, and cohort-based intelligence
Pahwa raised the debate many leaders have around small language models versus large language models. Naganathan responded with examples from both HERE and the partner-facing ecosystem OneUp.
He said the company mined its data built over more than 20 years and created anonymised cohorts. With a data lake and an AI engine on top, the company uses prompting to expose GenAI-driven insights to customers and partners.
For partners, he described use cases like renewal book insights and nudges that can improve outreach effectiveness. He gave examples of learning a customer base’s preference for WhatsApp over SMS, and even timing signals such as weekends between 4 pm and 6 pm being the right window to contact certain customers.
He also spoke about reducing unwanted outreach. If a customer typically renews only two days before expiry, he questioned the value of starting aggressive outreach two months in advance. Instead, he suggested a more precise cadence, such as reminders closer to the actual renewal behaviour.
Personalisation, but not the overreach version
Pahwa challenged the tension between hyper-personalisation and the risk-pooling foundation of insurance. Naganathan clarified that when he says personalisation, he does not mean overdoing it.
He described the historical context of tariffed products and limited standard covers, and argued that the move now is toward more precise offerings built around cohorts rather than one-size-fits-all products.
He also framed personalisation as giving customers freedom of choice. He spoke about transforming the core engine that supports the business to move from selling fixed products to offering modular covers, bundling them, and matching specific requirements.
Bias, hallucination, and responsible AI controls
Pahwa pushed on a critical area: bias in GenAI, especially when models are trained on western-dominant datasets. He asked whether HDFC ERGO had encountered statistically strong recommendations that were rejected due to bias, fairness, or reputational risks.
Naganathan said the company has had multiple instances and explained the approach as a combination of real-world testing, qualification, and refinement. He stressed that data quality is foundational, and in sensitive enterprise use cases the company prefers relying on enterprise data. He also indicated that for clear enterprise use cases, the company would rather work with smaller models trained on its own data, and in some cases host and run models in-house.
He described the need to distinguish where open datasets can be used and where they should not be used. He also pointed to stronger governance expectations in the market, noting that many organisations have evolved from InfoSec policies to cyber policies and now to AI policies that include governance sections and checks and balances.
He added that testing and QA have also changed, and GenAI can help developers and QA teams test scenarios and fine-tune outcomes across the lifecycle before rollouts.
The ROI question: AI is not cheap
Pahwa raised the reality that scrutiny on outcomes for CTOs and CMOs has increased, and asked if GenAI has made the job harder.
Naganathan said the bedrock remains the business case. He stressed that GenAI has a cost, including token costs, and leaders have to be careful where they apply the technology. He tied evaluation to per-transaction math and whether the spend makes meaningful sense against business value.
He also said he expects token costs to move downward over time, citing major investment interest in the data center space and the possibility that infrastructure scale will improve economics over the next couple of years.
Build versus buy and simplifying stacks
On the question of tool sprawl and stack complexity, Naganathan described it as a choice and framed it through the build versus buy lens. He argued there is now more focus on building, using proven models as components, rather than buying and integrating multiple products.
He shared an example of reimagining portal strategy about 18 months ago, introducing a new content management engine powered by AI, and building a lead scoring mechanism with a partner, also powered by AI.
Agents, empathy, and where bots should not replace humans
Pahwa asked when GenAI would shift from tools to semi-autonomous operators, including the idea of customers dealing only with AI agents.
Naganathan said parts of this are already happening, starting internally and gradually being rolled out to customers. He described a clear separation between simple, frequently asked requests like claim status checks or policy non-receipt, versus empathy-heavy situations such as accidents.
For emotional cases, he said customers do not want to speak to a bot, and the company uses sentiment analysis to route such interactions to experienced agents. He also described training human agents and building a twin to learn from top-performing contact center agents, with the intent of replicating excellence more consistently.
What is coming next
On near-term projects, Naganathan said HDFC ERGO has already rolled out multiple AI experiments, including use cases like summarising legal documents, quality checks, and voice bots.
Looking forward, he pointed to the agentic space, especially in retail health claims. He said early results are promising for deploying agents in low-level tasks and the company is exploring taking these to production in the coming months.
He also referred to auto claims, where image-based assessments have been used for a long time but often as an input to surveyors. For low-value claims, he said the company is evaluating an agent-driven approach that could create a different customer experience.
Closing note
The discussion reinforced a clear theme. GenAI is pushing the CTO role deeper into business strategy, but in insurance, the most meaningful wins will be measured in the moments that matter most to customers, especially claims. And even as automation expands, Naganathan repeatedly returned to one idea: separating what can be efficiently automated from what still requires human empathy, judgment, and trust-building.
Related Articles
