General Insurance Corporation of India (GIC Re) is building an in-house artificial intelligence assistant that is expected to help make faster and better-informed decisions, chairman and managing director Hitesh Joshi said in an interview. As part of its AI projects, the reinsurer is building an AI-powered assistant called Ragini (Retrieval Augmented Generation for Intelligent Natural Insights). Joshi, a three-decade GIC Re veteran who took the top job on 16 June after eight months as interim chief, has framed his tenure around one idea: resilience. He is building this through AI adoption, a rebalanced risk portfolio, a stronger credit rating and a phased return to overseas markets the corporation lost after its 2020 downgrade.

Edited excerpts:

You've just taken over after three decades at GIC. What are your immediate priorities?

I'd divide them into short-term and long-term. My time as executive director for reinsurance coincided with the rating downgrade—from July 2020 to the upgrade in October 2024—and that forced a lot of soul-searching.

We reworked our underwriting processes and focused on exposure management and portfolio rebalancing.

A reinsurer's stability rests on price adequacy, exposure management, catastrophe modelling and portfolio rebalancing. Put those together, and you optimize profitability against volatility, which ultimately determines the return we generate for shareholders.

When you talk about rebalancing the portfolio, what are you balancing?

Several things. Within property, it's catastrophe versus non-catastrophe exposure. Earthquakes and floods were once the major perils, but hurricanes and winter storms are now outpacing them as ocean temperatures rise. Beyond that, we balance property against casualty, non-life against life, and commoditised classes against specialty lines like marine, energy, liability, credit and agriculture.

Property is the bread-and-butter of the reinsurance industry, almost everyone underwrites it, so over time it becomes a commodity. The specialty classes behave differently, and the capacity there can be materially better depending on the cycle.

You have seven modellers with plans to increase it. Which areas would these be in?

Essentially property. We have seven certified catastrophe modellers today and plan to add another 17. Every property underwriter should understand modelling, because it's the foundation of catastrophe pricing.

This matters more as climate change scrambles the trends underwriters rely on. Dubai and South Africa are seeing floods in territories that weren't expected to flood, and hurricane behaviour is changing non-linearly. A building designed to withstand a wind speed of 230 kmph might simply disappear at 250.

Would you be using AI in this?

We've started a couple of projects and plan to scale up. One is called Ragini (short for Retrieval Augmented Generation for Intelligent Natural Insights). It's an AI-powered assistant that will let everyone in the organization query company data and get meaningful insights, while keeping official data private within our own infrastructure. It will act as a single portal across standardized documents in our IT systems and help leadership make faster, better-informed decisions.

The logic is simple. If 100 underwriters have each written 100 risks, that's 10,000 risk analyses. If that knowledge is documented and AI can mine it, the expertise stays with us even after an underwriter leaves. Every underwriter has different perceptions and biases, so once these are integrated, AI can bridge those gaps and build a kind of collective intelligence.

Why keep it in-house?

Why would we give it outside? There's an algorithm, and there's underlying data, and that data is very expensive and valuable, so it stays with us. We've already put safeguards in place—we've removed access to ChatGPT and licensed Copilot instead, because if every officer with an internet connection can paste data into a public tool, that data goes out. We can't take that risk. Right now, Ragini is a small pilot, but it's very much part of building resilience.

What is attrition looking like, and where are you losing people?

Our attrition remains low, at approximately 1.9% per annum. From the available data, most of those who leave move to other public-sector undertakings, primarily to take up postings nearer their hometowns. So it isn't a case of talent haemorrhaging to private competitors or GIFT City syndicates. My approach, in any case, isn't to cling to talent—it's to generate more of it than we lose. We depute people to our London and Malaysia branches and South African subsidiary, and AI reinforces the same goal: it safeguards against talent loss and multiplies the intelligence we already have.

Has your exposure to data centres and green assets grown?

Right now, those exposures are largely absorbed within the companies' own portfolios. If a data centre is covered within, say, a ₹1,000 crore capacity, it doesn't come to us; only when it crosses that threshold does it reach the facultative market. It's an emerging class.

The same can be said of the huge wind and solar farms in Gujarat and Rajasthan, large exposures, but the base is still small, so it isn't moving the needle yet.

Parametric, surety and cyber are focus areas that will grow off a small base over the next three to seven years.

The government divested part of its stake. How much is still to be sold, and how was the offer received?

We confirm that 2.4% of the government's holding is still to be sold, to ensure the minimum public shareholding of 25%. The recent offer for sale was oversubscribed 1.4 times overall, including retail. Bids came in for 12.285 crore shares against an offer of 8.772 crore shares, which included a greenshoe option of 5.263 crore shares. On the timeline and pricing for the remaining 2.4%, the corporation can't comment; the government is the competent authority to decide that.

You want to take overseas business back towards 40%. Over what period?

Over the next five to seven years, we plan to increase our overseas business from 25% to 40%. It'll be a challenge because the domestic market continues to grow, but we have sufficient solvency, which gives us ample room to grow. Practically, it means inching back into markets like Japan, where I can't simply walk back in at my old 4% share, I'll re-enter at 1% or 1.5% and rebuild the relationship gradually.

What about raising the share of life reinsurance?

We want to increase life's contribution significantly, from just under 5% today, moving closer to the mix seen among leading global reinsurers. But there are challenges, given the size of the Indian life market. We expect this to unfold gradually over the coming years, rather than to a fixed deadline.

With more reinsurers in GIFT City, how do you see the competition?

Around a dozen and a half players have come in, targeting India. Everest, Kuwaiti and Saudi reinsurers, among them, were already writing India from their headquarters, so capacity has largely moved closer to the market rather than being new. If they scale up, that can add competition.

But reinsurance is a subscription market largely written at a single price, and because global financial markets are integrated and every player must service its capital, it corrects itself. For domestic reinsurers, competitive strength depends on the capital deployed within the regulated entity, not the promoter's financial strength.

General Insurance Corporation of India (GIC Re) is building an in-house artificial intelligence assistant that is expected to help make faster and better-informed decisions, chairman and managing director Hitesh Joshi said in an interview. As part of its AI projects, the reinsurer is building an AI-powered assistant called Ragini (Retrieval Augmented Generation for Intelligent Natural Insights). Joshi, a three-decade GIC Re veteran who took the top job on 16 June after eight months as interim chief, has framed his tenure around one idea: resilience. He is building this through AI adoption, a rebalanced risk portfolio, a stronger credit rating and a phased return to overseas markets the corporation lost after its 2020 downgrade.

Edited excerpts:

You've just taken over after three decades at GIC. What are your immediate priorities?

I'd divide them into short-term and long-term. My time as executive director for reinsurance coincided with the rating downgrade—from July 2020 to the upgrade in October 2024—and that forced a lot of soul-searching.

We reworked our underwriting processes and focused on exposure management and portfolio rebalancing.

A reinsurer's stability rests on price adequacy, exposure management, catastrophe modelling and portfolio rebalancing. Put those together, and you optimize profitability against volatility, which ultimately determines the return we generate for shareholders.

When you talk about rebalancing the portfolio, what are you balancing?

Several things. Within property, it's catastrophe versus non-catastrophe exposure. Earthquakes and floods were once the major perils, but hurricanes and winter storms are now outpacing them as ocean temperatures rise. Beyond that, we balance property against casualty, non-life against life, and commoditised classes against specialty lines like marine, energy, liability, credit and agriculture.

Property is the bread-and-butter of the reinsurance industry, almost everyone underwrites it, so over time it becomes a commodity. The specialty classes behave differently, and the capacity there can be materially better depending on the cycle.

You have seven modellers with plans to increase it. Which areas would these be in?

Essentially property. We have seven certified catastrophe modellers today and plan to add another 17. Every property underwriter should understand modelling, because it's the foundation of catastrophe pricing.

This matters more as climate change scrambles the trends underwriters rely on. Dubai and South Africa are seeing floods in territories that weren't expected to flood, and hurricane behaviour is changing non-linearly. A building designed to withstand a wind speed of 230 kmph might simply disappear at 250.

Would you be using AI in this?

We've started a couple of projects and plan to scale up. One is called Ragini (short for Retrieval Augmented Generation for Intelligent Natural Insights). It's an AI-powered assistant that will let everyone in the organization query company data and get meaningful insights, while keeping official data private within our own infrastructure. It will act as a single portal across standardized documents in our IT systems and help leadership make faster, better-informed decisions.

The logic is simple. If 100 underwriters have each written 100 risks, that's 10,000 risk analyses. If that knowledge is documented and AI can mine it, the expertise stays with us even after an underwriter leaves. Every underwriter has different perceptions and biases, so once these are integrated, AI can bridge those gaps and build a kind of collective intelligence.

Why keep it in-house?

Why would we give it outside? There's an algorithm, and there's underlying data, and that data is very expensive and valuable, so it stays with us. We've already put safeguards in place—we've removed access to ChatGPT and licensed Copilot instead, because if every officer with an internet connection can paste data into a public tool, that data goes out. We can't take that risk. Right now, Ragini is a small pilot, but it's very much part of building resilience.

What is attrition looking like, and where are you losing people?

Our attrition remains low, at approximately 1.9% per annum. From the available data, most of those who leave move to other public-sector undertakings, primarily to take up postings nearer their hometowns. So it isn't a case of talent haemorrhaging to private competitors or GIFT City syndicates. My approach, in any case, isn't to cling to talent—it's to generate more of it than we lose. We depute people to our London and Malaysia branches and South African subsidiary, and AI reinforces the same goal: it safeguards against talent loss and multiplies the intelligence we already have.

Has your exposure to data centres and green assets grown?

Right now, those exposures are largely absorbed within the companies' own portfolios. If a data centre is covered within, say, a ₹1,000 crore capacity, it doesn't come to us; only when it crosses that threshold does it reach the facultative market. It's an emerging class.

The same can be said of the huge wind and solar farms in Gujarat and Rajasthan, large exposures, but the base is still small, so it isn't moving the needle yet.

Parametric, surety and cyber are focus areas that will grow off a small base over the next three to seven years.

The government divested part of its stake. How much is still to be sold, and how was the offer received?

We confirm that 2.4% of the government's holding is still to be sold, to ensure the minimum public shareholding of 25%. The recent offer for sale was oversubscribed 1.4 times overall, including retail. Bids came in for 12.285 crore shares against an offer of 8.772 crore shares, which included a greenshoe option of 5.263 crore shares. On the timeline and pricing for the remaining 2.4%, the corporation can't comment; the government is the competent authority to decide that.

You want to take overseas business back towards 40%. Over what period?

Over the next five to seven years, we plan to increase our overseas business from 25% to 40%. It'll be a challenge because the domestic market continues to grow, but we have sufficient solvency, which gives us ample room to grow. Practically, it means inching back into markets like Japan, where I can't simply walk back in at my old 4% share, I'll re-enter at 1% or 1.5% and rebuild the relationship gradually.

What about raising the share of life reinsurance?

We want to increase life's contribution significantly, from just under 5% today, moving closer to the mix seen among leading global reinsurers. But there are challenges, given the size of the Indian life market. We expect this to unfold gradually over the coming years, rather than to a fixed deadline.

With more reinsurers in GIFT City, how do you see the competition?

Around a dozen and a half players have come in, targeting India. Everest, Kuwaiti and Saudi reinsurers, among them, were already writing India from their headquarters, so capacity has largely moved closer to the market rather than being new. If they scale up, that can add competition.

But reinsurance is a subscription market largely written at a single price, and because global financial markets are integrated and every player must service its capital, it corrects itself. For domestic reinsurers, competitive strength depends on the capital deployed within the regulated entity, not the promoter's financial strength.