Page 41 - The Insurance Times February 2026

Insurance AI

The Missing

Architecture of

Insurance AI: Why

India Must Shift From Sandeep Khuperkar

CEO & Founder
Use Cases to Systems DSW Unify AI

India's insurance sector is at a pivotal point. Digital adoption is strong, regulators are encouraging
innovation, and customer expectations are rising faster than product cycles. But the real
breakthrough we need next will not come from "more models." It will come from re-engineering the
systems in which AI operates.

F or some time now, the insurance industry has adopted Y A fraud score triggers flags but doesn't integrate with

claims approval systems.
AI in fragments-a fraud model here, a claims bot there,
and an underwriting score somewhere else. Most of
these initiatives succeed in isolation. Yet, when insurers Y Underwriting recommendations do not align with policy
rules or pricing engines.
attempt to scale them across a business line or a policy
lifecycle, they confront a surprising bottleneck: the problem Y Audit trails are inconsistent because each tool records
is not the model; the problem is the system around it. decisions differently.
Y Scaling across states, regions, or business lines creates
India's insurance sector is at a pivotal point. Digital adoption inconsistencies.
is strong, regulators are encouraging innovation, and
customer expectations are rising faster than product cycles. These issues are not mathematical failures; they are
But the real breakthrough we need next will not come from architectural ones. Insurers don't lack algorithms. They
"more models." It will come from re-engineering the systems need systems that coordinate algorithms across the
in which AI operates. enterprise. And this is exactly where India's next leap in AI-
enabled insurance will be defined. The System Lens: Why
This shift from task automation to system orchestration is Insurance Must Think Beyond Just Models. Insurance is a
where the next decade of InsurTech innovation will unfold. system business, not a task business.

AI Works… Until It Has to Work With A claim settlement is not a single AI event. It is a series of
tightly coupled steps:
Everything Else 1. Intake and documentation
Ask any insurer why AI pilots stall. The answers sound similar 2. Policy validation
across motor, health, life, and commercial lines:
Y The model works, but the workflow doesn't accept its 3. Coverage interpretation
output. 4. Fraud risk scoring

36 February 2026 The Insurance Times