Agentic AI has been heralded as a top tool for efficiently orchestrating customer engagement activities at insurance companies.
It can:
- Automate repeatable, rule-based processes that affect customer experience (think claims processing and customer onboarding).
- Rapidly retrieve disparate data, analyze risks and recommend decisions.
- Help lower operational costs while keeping humans focused on complex cases that require empathy and judgment.
But there’s a fine line between saving time and reducing costs versus keeping customers happy and engaged. Especially for insurance companies.
Read the other blogs in this series about agentic AI in insurance
Insurance – an emotional, trust-based business
Consider the emotional nature of the services customers need from insurers. They typically engage with providers:
- When they want coverage for a valued asset, such as a home, a car, or a diamond ring.
- To ensure they’re sufficiently covered for their health or life insurance.
- If they experience a sudden change of circumstances, such as a death in the family.
- After suffering some other type of loss, like a fallen tree that damages their roof.
Understanding insurance customer engagement
For insurers, engaging customers involves attracting new customers, assisting applicants through the insurance process, retaining existing customers and appropriately cross-selling new products. Engagement often extends to groups – like employees in a group insurance policy, or to families, as a unit.
But dissatisfaction from one person in a group can have an outsized influence on how others respond. So, if a father is upset about how his auto accident claim was handled, he is likely to leave his insurance carrier – along with the rest of that group (his family). Imagine the potential impact if someone covered by an employer’s group insurance policy became highly dissatisfied.
From in-person to digital
Insurance was traditionally a face-to-face business, with regular in-person touchpoints between the human insurance agent and the insured. Such check-ins tended to keep things in balance between customer expectations and the insurer’s coverage.
But today’s customers expect to use digital channels to compare offers and report claims. Their expectations for fast, reliable digital service are high, increasing competition and complexity of the business.
Some insurers have gone fully digital, with no physical location or phone support. Their aim is to reduce costs while maintaining customer satisfaction. This digital shift poses serious challenges in preserving the human touch – especially important for complex or trust-sensitive products like life insurance.
“By using machine learning and deep learning, we are significantly improving the experience for the hybrid customer. SAS helps us efficiently approach our target group with relevant information and offers in real time. While AI is a relatively new addition to our analysis toolkit, we’ve already seen the real value it can bring – and we know that AI will play a crucial role in helping us become a digital leader in the insurance sector.”
— Tim Stettner, Head of Advanced Analytics at ERGO
Trusted AI and human judgment
Insurance is a trust business, as people purchasing coverage entrust their lives and livelihoods to their insurance carriers. Likewise, customers’ decisions often come down to trust in the carrier, the coverage, and the need for insurance in the first place.
Given the starring role trust plays with insurance customer engagement, it follows that trust in AI is non-negotiable for insurers wishing to use tools like agentic AI to retain and grow their customer base.
Human intelligence and AI
Humans certainly have a grave responsibility when it comes to training the models that guide AI agents in carrying out detailed processes across the customer life cycle. But it’s tricky to capture this in an AI model. There are many nuances that humans routinely consider in working with individual customers. Their knowledge is based on years of personal experience engaging with and counseling policyholders.
Is it feasible for AI agents to perform duties historically handled by empathetic people?
For the best results with customer engagement, insurers need to:
- Figure out when it’s best to keep human agents involved in personal interactions with customers versus turning to AI.
- Incorporate the wisdom of experienced human agents into the AI models that run customer engagement and related activities.
- Recognize the degree to which knowledge gaps will affect model quality – and use AI agents accordingly. Also, keep in mind that legal and regional differences complicate AI use for insurance, especially in the areas of pricing and contract management.
How do GenAI and AI agents work in customer engagement?
Many insurers use generative AI (GenAI), including large language models (LLMs) and chatbots, to provide 24×7 customer service. But even sophisticated chatbots can struggle with the complexity and emotional nature of insurance claims, potentially leaving customers disillusioned.
Agentic AI can work in conjunction with GenAI and chatbots to automate repeatable, rule-based processes, like those involving claims. Operating behind the scenes, AI agents can rapidly access and integrate data from multiple sources and legacy systems to streamline workflows. Before establishing this workflow, it’s important for insurers to unify their siloed systems so they can train their AI agents to check all the right sources of information.
If this chatbot-AI agent process delivers on customer expectations for quick, reliable service, the technology can streamline processes, reduce operational costs and improve customer satisfaction. But if it fails to deliver, the insurer’s reputation is damaged – potentially causing the insurer to reverse course and return to a call center model.
Bottom line: insurers must diligently train their AI models and carefully select the tasks they allow AI to perform for customers.
Models work like a decision tree, of sorts, based on sets of business rules. The model has access to all the rules (and data), and it learns automatically. As they absorb data, AI models learn to predict outcomes or solve specific tasks without explicit instructions at every step. And they become more refined as they get more data.
Example: Car accident
Let’s say you have a car accident at 3:00 a.m. You immediately report it to your insurer via a phone app (which involves talking with a chatbot). The chatbot can handle the easy parts by asking: What happened? Who’s involved? Which vehicle was it? Do you need to call the police, get a tow truck, or call an ambulance?”
The answers to these “Yes/No” questions trigger a next action, like calling for a backup car, the police, etc. These steps could be done automatically by an AI agent. Simultaneously, an AI agent could retrieve information about the insured person and check the details of their insurance policy coverage in the background.
If everything is routine, this interaction may provide the customer with everything they need for now.
But consider a chain-reaction crash with three cars, conflicting statements and injuries. An AI agent can capture details and trigger next steps – towing, rental and medical triage. But liability isn’t a “Yes/No” decision, and customers want to feel heard when the outcome affects their finances and safety.
That’s when a human adjuster needs to join: to investigate, weigh evidence and communicate decisions with empathy. When the claim becomes a dispute, a gray area or a high-emotion moment, the AI agent should pause and bring in a human.
Example: Obtaining insurance coverage
Insurance companies tailor customer engagement based on the product’s complexity and the individual customer’s needs. So digital experience should vary by product type. Consider:
- Simple products like travel insurance. This type of coverage is essentially a short-term commodity, so it’s suited to digital-only interactions. Many customers will gladly opt in for coverage simply by checking off a box when they book a trip, for example.
- Customized products. Sensitive insurance products – such as life or health insurance – require detailed personal knowledge about the customer. Because of the nature of the information shared and the need for highly tailored products, expert human knowledge is required. Generally, for complex insurance products like this that depend on human empathy, judgment and trust, it’s best to keep experienced human agents involved. The more tailored to individual needs, the more sophisticated the customer service should be.
The risks of not getting customer engagement right
Low or slipping customer engagement can lead to customer dissatisfaction and cancellations. Insurers balance many competing factors to keep customers happy while still meeting internal risk thresholds. For example:
- Insurers expect to have appropriate contact from and engagement with customers – but it’s a balance. Excessive contact initiated by the customer may indicate a high risk or even fraud, while too little may suggest the customer is considering switching insurers.
- They need to maintain routine engagement and cross-sell new insurance products, but they must be careful not to annoy or overwhelm customers.
- Insurers want to provide coverage to customers, but they need to carefully evaluate the risks involved. Finding the right balance of risk in the customer pool requires excellent data quality, connected backend systems across functions, and effective risk assessment techniques.
- Country-specific regulatory compliance also plays a role in deciding when and how to engage with customers. It plays a particularly crucial role for multinational carriers.
Maintaining customer trust and managing emotional interactions are critical for insurers today. And AI cannot fully replicate human intelligence or empathy. The risk of misinformation from poorly trained AI models underlines the need for careful governance, human oversight and transparency.
It all comes down to the customer
As they balance AI technologies and human interactions, insurers must carefully navigate cost pressures, regulatory constraints and customer demands.
Along the way, never forget: The customer is an insurer’s most valuable asset. Any new technology, including agentic AI, needs to be implemented thoughtfully – and with guardrails like AI governance – to preserve trust and quality of the service.
