The insurance industry’s AI journey did not begin with generative AI.
Long before boardrooms became consumed with conversations around large language models, insurers were grappling with a more fundamental challenge: modernising decades-old technology foundations that were never designed for a digital-first world.
For Durgesh Nandan, CIO at IFFCO-TOKIO General Insurance, the industry’s most significant transformation is not AI itself. It is the shift from legacy, monolithic architectures to modern, intelligent platforms capable of supporting automation, analytics, AI, and real-time decision-making at scale.
“Most insurers started their digital journeys years ago, but many were still operating on legacy architectures,” says Nandan. “The industry is now moving from basic digitisation to modern digitisation, where AI, analytics, workflow automation, and intelligent decision-making can become part of the core operating model.”
That shift is driving one of the largest technology modernisation cycles the insurance industry has seen in decades.
Building the foundation for intelligent insurance
Across the industry, insurers are replacing aging core systems that were originally built on technologies designed for a different era. IFFCO-TOKIO is no exception.
The company is currently undertaking a large-scale core transformation program with TCS, aimed at creating a modern technology foundation capable of supporting future AI, automation, analytics, and ecosystem-driven innovation.
The objective is not simply to replace technology. It is creating an architecture that allows innovation to scale.
“AI should not become something that needs to be rebuilt every time the core platform changes,” says Nandan. “The idea is to create plug-and-play capabilities that remain relevant even as the core system evolves.”
That philosophy has shaped the company’s AI strategy. Rather than waiting for its multi-year transformation program to conclude, the company has begun deploying AI-led use cases that can integrate seamlessly into both existing and future environments.
The result is a technology roadmap where modernisation and innovation move in parallel.
When AI solves real problems
One of the biggest misconceptions surrounding AI, according to Nandan, is that every process should be automated simply because the technology exists.
Successful AI adoption starts with identifying business problems where intelligence can create a measurable impact.
Health claims processing offered one such opportunity. Under regulatory guidelines, insurers must often respond to hospitals within strict turnaround times regarding claim admissibility and approvals. A single claim can involve 50 to 60 documents, making manual review both time-consuming and resource-intensive.
IFFCO-TOKIO implemented an AI-powered claims solution that uses OCR, machine learning, and intelligent document processing to analyse incoming documents, identify discrepancies, summarise critical information, and flag potential anomalies.
“What previously required someone to review 50 or 60 documents manually can now be summarised in a couple of minutes,” Nandan explains. “The system also identifies inconsistencies proactively rather than waiting for them to be discovered later.”
The impact has been significant. The company currently processes approximately 5,000 to 6,000 health claims every month through AI-assisted workflows. After initially routing only 30% of claims through the AI process, it expanded adoption to 100% since March 2026, within a few months after validating the results.
Motor claims present another compelling use case. With image manipulation and deepfake technologies becoming increasingly sophisticated, insurers face growing challenges around fraud detection. To address this, the company has deployed AI models capable of detecting edited images, duplicate submissions, and potential anomalies before claims are processed.
The company has also developed several AI-powered capabilities in-house, supported by GPU-based infrastructure and internally developed machine learning models.
Reimagining risk through AI
Perhaps one of the most interesting examples of AI-driven transformation at IFFCO-TOKIO is iDrishti 2.0, the company’s internally developed inspection platform.
Traditionally, risk inspections depended heavily on the individual expertise and judgement of surveyors. Two inspectors evaluating the same site could potentially focus on different aspects of risk, leading to inconsistencies in underwriting decisions.
iDrishti 2.0 changes that dynamic. The mobile application guides inspectors through a structured and dynamic assessment process, automatically generating subsequent questions based on previous responses. Images, videos, documents, and observations are captured through a standardised workflow before being converted into a detailed risk assessment report.
“What we achieved was standardisation,” says Nandan. “The inspection process is no longer dependent on who conducts it. The system ensures that every relevant aspect of risk is evaluated.”
The platform ultimately generates a risk score and recommendations that allow underwriters to make faster and more consistent decisions.
The broader implication is significant: AI is gradually moving insurance from retrospective assessment toward more structured, predictive, and data-driven risk evaluation.
The ROI question
Despite the excitement surrounding AI, Nandan believes the industry is entering a more mature phase of adoption.
“The first phase was about the buzz,” he says. “Now the conversation is shifting towards ROI.”
His perspective is shaped by experience. Several years ago, IFFCO-TOKIO invested heavily in an internally hosted AI-based motor claims assessment platform. The company purchased dedicated hardware, trained machine learning models using large image repositories, and built sophisticated capabilities for damage assessment.
However, maintaining infrastructure, continuously retraining models for new vehicle designs, and managing ongoing operational complexity proved challenging. “If the use case is not suitable, or if infrastructure costs are not optimised, ROI will become a problem,” Nandan says.
Today, the company follows a pragmatic approach, often favouring transaction-based consumption models where partners absorb infrastructure and model maintenance costs.
For Nandan, AI success depends on three questions: Is the use case suitable? Can the outcome be measured? And does the economics make sense?
Why human judgement still matters
As Agentic AI gains momentum, many enterprises are exploring workflows that require little or no human intervention.
Nandan believes that the future is approaching—but selectively.
“There are use cases where agentic systems can operate independently,” he says. “But only where the risk of occasional errors is acceptable and where financial or business impact remains limited.”
The distinction is important. Tasks such as extracting information from documents or validating structured inputs may be highly suitable for autonomous processing. Mission-critical financial decisions, however, still require human oversight.
“You should not choose the wrong use case and then blame AI,” he says. “You need to understand both the input and the expected output before deciding how much autonomy to provide.”
The next phase of transformation
Looking ahead, IFFCO-TOKIO’s priorities are less about adding new technologies and more about simplifying complexity.
The company’s long-term vision centres on a lightweight core platform supported by a unified digital layer for customers, partners, underwriters, claims teams, and business users.
Rather than maintaining fragmented applications and disconnected workflows, the goal is a guided, intuitive experience in which systems lead users through processes while AI and automation operate invisibly in the background.
“Transformation is ultimately about simplification,” says Nandan. “The architecture must be simple, scalable, and intelligent enough to guide the user rather than forcing the user to navigate complexity.”
That philosophy may ultimately define the next chapter of insurance technology.
The future will not belong to organisations that deploy the most AI. It will belong to those who know where AI creates value, where human judgement remains essential, and how to bring both together on a modern digital foundation.