Page 72 - IMDR Journal 2025

Research Article
● Fraud Detection AI systems are highly effective at
detecting fraud, with an 85% detection rate compared to
60% for traditional methods. This helps insurers reduce
financial losses and improve operational efficiency.
However, AI-driven systems face challenges in terms of
transparency, fairness, and customer interaction. While AI is
efficient, it often lacks the human touch that customers
value, particularly in sensitive situations. Additionally, the Fig 6.1.2
lack of transparency in AI decision-making can erode trust,
particularly when claims are rejected without clear 5. Digital Literacy Challenges
explanations.In contrast, human-driven systems offer o Urban respondents (66.7%) reported higher satisfaction
greater transparency and empathy, but are slower and more with AI systems compared to rural respondents (33.3%),
prone to errors. A hybrid model, where AI handles routine highlighting India's digital divide.
tasks and humans review complex cases, may offer the best
of both worlds, balancing efficiency with empathy and trust. Major Consumer Concerns
● Lack of clarity in AI decision-making ("black-box"
systems).
FINDINGS
● Fear of algorithmic bias affecting claim approvals.
The study reveals critical insights into consumer
● Impersonal customer interactions with chatbots.
perceptions of AI-driven claims processing in the Indian
health insurance sector. Key findings include:
1. Efficiency vs. Trust RECOMMENDATIONS
o 92.6% of respondents perceive AI-driven systems as If insurance companies want people to trust digital claim
more efficient than traditional methods, with claims systems, they need to do more than just be fast they need to
resolved in 2-3 days compared to 10-15 days manually. be clear, fair, and people-focused. Here are some ways they
o However, only 45.1% expressed trust in AI decisions, can make that happen:
citing concerns about transparency and fairness. This 1. Be Transparent and Easy to Understand
highlights a gap between the efficiency of AI and consumer Use simple language when explaining claim decisions. For
trust in its decision-making processes.
example: “Your claim was denied because we didn’t receive
2. Transparency Gap enough medical documents.” Create dashboards or mobile
o 54.9% of respondents reported experiencing claim apps where customers can easily check their claim status and
rejections without clear explanations, leading to frustration see how decisions are made step by step.
and mistrust. For example, one interviewee noted, "My 2. Make Sure It’s Fair for Everyone
claim was rejected without a clear reason. It felt arbitrary."
Check whether the system treats all people fairly, regardless
o This lack of transparency is a significant barrier to trust, of their age, income, or where they live. Use real-world data
particularly in a market like India where digital literacy from across India cities, villages, different regions so the
varies widely. system understands everyone better. Ask outside experts to
review the system for fairness and follow ethical standards.
3. Keep the Human Touch Where It Matters
Let technology handle the simple stuff like checking
documents or confirming details. But bring in real people for
more complicated claims or when someone wants to appeal
a decision. For instance, Lemonade, a global insurer, uses a
chatbot that passes cases to a human whenever empathy is
needed.
Fig 6.1.1
4. Communicate Clearly With Customers
3. Perceived Bias Tell people how these tools work using short videos, FAQs,
o 29.4% of respondents reported suspicions of bias, or even short workshops. ICICI Lombard’s InstaSpect app is
particularly in health insurance claims. Older policyholders a good example it shows how a photo of car damage is used
and rural customers felt disproportionately affected by AI to process claims quickly. When a claim is denied, explain
rejections, suggesting that AI systems may inherit biases why clearly and step by step so the customer isn’t left
from training data. confused.
4. Hybrid Preference 5. Ask for Feedback And Use It
o 68.6% of respondents preferred a hybrid model where AI Give customers a way to challenge decisions or give
handles initial processing, but humans review complex or feedback. If many people point out the same problem, use
rejected claims. This approach was seen as balancing that information to fix the system. Regularly ask people how
efficiency with empathy and trust. they feel about the process this helps improve the service

63