conversational ai insurance

The almost 300-year old insurance industry has been relatively slow to react to the digital age. But with the rapid pace of technological innovation and changing customer expectations in recent years is starting to change things. 

 

Artificial intelligence lies at the frontier of technologies that could disrupt the insurance landscape.   New InsurTech startups and Tech incumbents are already putting forth innovative use cases, and even traditional insurance players are being forced to follow suit. In fact, a 2015 study conducted by IBM showed that 95% of insurance executives were intending to start or continue investing in AI capabilities in the future.

 We have put together this guide to give you a comprehensive understanding of conversational AI in insurance  – from the various use cases, how they work, how to plan their implementation in your organisation and what further advancements we can expect in the future.

Contents:

  1. AI Use Cases in Insurance
  2. Types of Insurance
  3. Conversational AI Uses Cases in Insurance
  4. Unique Challenges for Conversational AI in Insurance
  5. What Insurance Providers Should Know
  6. The Future of Conversational AI in Insurance
 


 

AI Use Cases in Insurance

In general, artificial intelligence can be applied to a the insurance value chain via a number of entry points. Below are just a few of these use cases.

 

ai uses cases insurance-1

 

  • Product Development  

Text analytics and natural language processing can be applied to scan and structure existing policies and product descriptions to develop future products faster and more efficiently. Specific AI models in pattern and anomaly detection as well as speech recognition can analyse customer feedback for patterns to create more targeted products or to improve future products.   

 

  • Marketing and Sales  

Customer behaviour on support calls and social media posts provide valuable signals for marketing and sales. NLP models can analyse these interactions to develop new marketing campaigns. These along with voice recognition techniques can also detect emotions in customer speech to improve personalisation. 

 

  • Underwriting and risk-rating  

Fraud detection is another key use case of AI in insurance. Specifically, text analytics using NLP can scan for ambiguities and rate risks in insurance applications based on claims. Risk assessment can also be made more precise by predicting premiums based on past risk assessments. 

 

  • Customer servicing   

Insurance providers invest a lot of resources in customer support to keep their policy holders informed and satisfied. AI tools employing NLP can help save providers time and cost by generating structured feedback based on customer requests, while past answers can be analysed to match customer enquires to automate some of the common questions. Voice, image and video recognition can be used to improve customer communications by detecting their emotions and sentiments.   

 

  • Claims management   

By analysing historical claims reports, AI can generate structured data sets and templates for new claims to improve efficiency and speed up processing.   

 

  • Financial assets   

NLP solutions can help insurance policy holders make critical decisions on their portfolios by analysing news and social media data to detect market movements and anomalies faster. Robo advisory solutions which analyse market movements to help investors forecast accurately are also becoming popular.  

 

  • Operations  

Within the insurance firm, AI solutions can help improve business operations in a number of ways. NLP models can transform process descriptions into structured data, recommendation engines can identify suitable candidates for hiring based on their social network data while facial recognition technology can be used to monitor security cam and warn employees about potential workplace safety issues.   


Back to contents


 

Types of Insurance Products

The insurance industry is by no means one with a single product or service. There are multiple types of insurance products, with various stakeholders in each as well as a variety of distribution channels. 

 

 

Category  

What it Covers  

Key Stakeholders   

Distribution Channels  

Health Insurance  

Medical expenses in for illnesses or accident  

Insurance provider  

Corporate HR teams  

Policy holders  

Corporate Employees  

Agents 

Consultants 

Brokers 

Self-service  

Life Insurance  

Financial support for dependents in the event of death or disability  

Savings and Investment  

Retirement income  

Insurance provider  

Policy holders  

Agents  

Consultants  

Agencies (employing a workforce of agents)  

Independent brokers  

Banks and financial institutions  

General Insurance  

To protect assets like car, property and for covering travel  

Insurance provider 
Policy holders 

Self-service 

Brokers 

Agents 

 

Insurance products can also be classified broadly based on how homogenous they are. This refers to how simple the product is and how much historical data on claims is available for providers to assess risks. General insurance and travel insurance falls under this category. Because of their nature, these are also products which users can research about online, aggregate options and compare prices, lending themselves well to self-service options.   

 

Products like health and life insurance on the other hand can be more complicated, covering different scenarios, demographics and uses. Life insurance could be relevant for young couples planning to save for their children’s future, investing in various savings schemes while pre-retires could have products specific to retirement income. Health insurance too can vary depending on pre-existing conditions, specific coverage for critical illnesses and the needs of employees in an organization.  

 

These products are also less homogenous in that there is a relative lack of historical claims data that insurers can use to predict future losses. For these complex products, the general practice is for users to go through an agent who acts as an intermediate advisor.  

Back to contents


 

Conversational AI Use Cases in Insurance

The specific application of AI in chatbots and conversational agents finds multiple use cases in insurance.   

 

Automating Repetitive Queries

Due to the diversity and complexity of insurance products, it is natural for customers to enquire about the finer details of their policies. In fact, even agents and the various stakeholders in distribution need their questions clarified. Put together, the volume of these queries can be massive for an insurance provider to handle, even with the help of large contact centres housing hundreds of customer service employees.   

 

Conversational AI solutions can automate at least the most repetitive and frequently asked questions among these. By providing a first level of support, these systems enable the support teams to focus on more important tasks that are value-adding to the customer and revenue-generating for the provider.   

automate chatbot queries
agent servicing bots

Education and awareness for agents servicing 

Agents are often the go-to resource for customers and policy holders to seek clarification about their products. And yet, often these agents themselves find it challenging to keep up with the details of the products they need to sell. The specifications of insurance policies, their pricing and premiums, coverage clauses, term limits, renewal processes and other finer details can change over time.  

 

To make matters more complicated, agents are also inundated with communications from providers on various aspects including promotions, monthly offers and discounts, announcements and more. They may simultaneously have to monitor their portals for notifications and more formal transactions, email dashboards for important communications as well.  

 

In such an information-heavy environment, it becomes challenging for agents to find the info they are looking for at a moment’s notice when interacting with customers. This could potentially affect their relationships with policy holders. Conversational AI solutions can help these agents access the info they need at their fingertips through simple queries via a conversational interface.  

 

Self-servicing for Policy Holders  

Insurance providers can also use conversational AI solutions to go directly to the customer. Through an intelligent chatbot embedded on company website or in an app, the policy holder can go through the entire insurance purchase journey themselves.  

 

This works most effectively for simpler types products where the features tend to be similar and easier to compare without the end user needing to possess much domain knowledge. Travel insurance, motor insurance and mutual funds are some such products where giving policy holders the option of self-servicing can be extremely beneficial 

 

For example, they can explore the various products available, compare prices and premiums, identify the best fit based on their profile and continue to purchase, all by asking questions of the bot. For these types of simpler products, insurance firms are finding it more beneficial to enable the end user with self-service options rather than owning a relationship through agents and brokers.  

insurance self servicing
insurance call centre

Boosting Call Centre Operations Efficiency

Conversational AI solutions can be a huge advantage for insurance firms which employ large teams of customer service representatives in contact centres. These contact centre agents are tasked to support customers through various channels including phone, email, messenger apps and chat portals. Phone calls on their own take much of their time, as they can only talk to one customer at a time.  

 

But with live chats, they may be able to handle two to four sessions concurrently. Conversational AI agents can help them scale their support by automating around 80 percent of these interactions, increasing their productivity by anywhere between 200-400 percent.  

 

The ideal situation would be for a conversational AI agent to fully automate at least the most repetitive queries. But to get there requires consistent ongoing training of the bot. This can also be made easy by having bots suggest the most appropriate responses for the support staff to click from and acknowledge during live chat. Not only does this help with the bots training, it also increases convenience for the staff and drastically improves the efficiency of the contact centre.  

 

Improving the Customer Journey 

Most large insurance providers today are exploring their digital transformation journeys. One of the big initiatives we find among insurance firms is the drive towards modernising their customer experience journey. With the multitude of channels available, resources employed to oversee the channels, and contact centres to manage everything, the cost of servicing can mount quite quickly, compromising ROI.   

10X guide Page content icons (2)

 

Read more about the various conversational AI use cases in insurance.


 Back to contents


 

Unique Challenges for Conversational AI in Insurance

The insurance industry has its own challenges when it comes to conversational AI implementation.  

 

  • Complexity of products 

It is well known that insurance products can be quite complex to understand. The different types of insurance products available, the right channels to go for purchasing them, quotes, premiums and riders can all be confusing for the customer. Very often, customers admit well after purchase that they are still unsure of the details of what they bought.   

 

Conversational AI solutions which can guide the customer through the purchase journey by providing them with clear information at every stage will earn their loyalty. In addition, this will also be an opportunity for providers to gain a competitive edge over others who may still be sticking to traditional acquisition and retention practices.   

 

  • Data collection challenges  

 

To get the most out of conversational AI, insurance providers need to train the system with a variety of different data sets. Data on company info, types of products, terms and conditions, exceptions and other publicly available data such as social media sentiments and financial market movements may be easily available.   

 

But to offer a personalised service to the customer, you need to combine this with their own data, which may be slightly more difficult. Most insurance products relate to customers’ demographics such as age, gender, life stages, family size, health conditions and historical purchase preferences. Customers may not be readily willing to disclose such information for obvious reasons and yet analysing such data in a conversation can help providers tailor their solutions more accurately. Thus, any conversational AI solution will need to come with a guarantee of privacy protection and secure data collection practices.   

 

  • Changing definitions and coverage 

 

Insurance products may be subject to revisions and redefinitions from time to time. We are always learning more and understanding the possibilities and limits of human life through scientific research. Medical advancements and health trends also impact the quality of life and monetary costs of injuries, diseases and other accidents.   

 

Any technology that relies on information about these aspects will need to be ready to reflect changes as and when they happen. Conversational AI is no exception.  

 

For example, consider the recent changes in the definition of critical illness as announced by the Life Insurance Association of Singapore. Deafness (Loss of Hearing)” was amended to "Deafness (Irreversible Loss of Hearing)", to take into account the possibility that hearing could be restored through treatment as and when better technology becomes available in the future. In a conversational AI interface, providers need to ensure that these changes are reflected when leads and customers enquire about health insurance. 

  

  • Sales force support and management 

In trying to cater to the end users and policy holders, insurance firms must also consider the needs of agents, brokers and the sales force. It is the agents who often own the relationship with the end user, especially for complex insurance products. And so it is in the interest of the insurance provider to support and manage the agents effectively.  

 

Some forward thinking insurance firms like AIA are already thinking of ways to help their agent workforce be more productive by enabling them with mobile apps and omnichannel experiences. But there is still more to be done to improve agent retention. 

 

The lack of post-sales service and support happens to be one of the major reasons why agents decide to end their relationship with the insurance provider. Beyond providing them with education about the products, they also need to be supported on aspects pertaining to commissions, payment terms and policies.  

 

With the rise of conversational AI and chatbots, there is also a growing concern among agents that their traditional roles of client servicing will be displaced. Insurance providers should allay their fears and educate them about the value of human agents in providing high level client servicing and that they will still expect sales revenue from human agents over direct purchases for complex products. 

 

  • Legacy systems in distribution channels 

 

The technology infrastructure at large traditional insurance firms were built to process large volumes of transactions. This meant they comprised mainframe computers and legacy systems. These may not have not been modernised yet or in many cases may have additional peripheral features tagged on over the course of their evolution to support new digital initiatives one by one.  

 

As a result may not connect with others, or may have different formats for storing data and in general create silos that limit innovations and new product developments aimed at improving customer experience. In fact, a recent PwC report on legacy insurance systems showed that on average it took six to nine months for insurance firms to develop and test a new product, costing between US$ 400,000 and US$ 900,000. 

 

The nature of the distribution of insurance products complicates things even further. There can be multiple channels dedicated to agents, brokers, bancassurance and customer service contact centres. Different agency workforces could also be using their own portals to communicate with the provider.  

 

Conversational AI systems have the potential to bring these disparate systems together in one unified omni-channel customer experience platform. But before such a solution can be implemented, the siloes must be broken through so that the legacy systems can interface with modern ones.  

Back to contents


 

What Insurance Providers Should Know About Conversational AI Implementation 

Due to the traditionally slow nature of insurance firms to adopt modern technologies, they will need to keep the following points in mind if they are considering conversational AI solutions.  

 

In some cases, there may also be a basic version of a chatbot or a proof of concept already in place, and the challenge would be to take it to production or scale it to meet the organisation’s needs.   

 

1. Prepare for Scaling from Proof of Concept to Production  

The first step towards implementing conversational AI systems often turns out to be a Proof of Concept. But this stage is relatively easy and can often be accomplished by an in-house team of developers, using an off-the-shelf framework. But scaling it to meet the true demands of a large insurance organisation, with their many distribution and customer service channels, can be a challenge on another level altogether.  

 

In fact, many enterprises make the mistake of taking it easy, only to realise that it takes more work than estimated. For example, in deciding what intents to answer, insurance providers will have to spend more time and resources during production. It could involve text mining or even transcription of call center data, in order to uncover possible intents, along with additional work required to cleanse and augment the data to a usable state. And in training the bot to recognise the intents, the NLP model might need to incorporate words and phrases specific to the insurance industry as well as phrases, misspellings, loanwords, shorthand, slang, unusual synonyms, market misconceptions and even emoticons. To ensure the project can be scaled effectively from PoC to production, companies will need to seek expertise from specialised vendors to  

 

  • Create a robust language model that is trained on the domain, company and the language of the users  
  • Combine as many data sources as possible to uncover the intents, and organise them in a sensible hierarchy  
  • Conduct a thorough content audit of the scripts and knowledge repositories, including ownership, vagueness and the work needed to transform knowledge into dialogue    
  • Perform an audit of the applications involved in transactions conducted by the chatbot to note its purpose, the state of its API, security considerations, and the complexity of dialogue required to support it. 

 

2. Connect legacy systems with API middle layers

Insurance companies differ in the extent to which they have adopted new technology solutions. Taking the next step towards conversational AI will require them to think about their existing infrastructure and how it can accommodate the chatbot solution.   

 

Unlike large finance or technology-heavy organisations, insurance firms have not had the need to transform their IT infrastructure in recent times. In general, they have adopted a mindset of “if it ain’t broke, don’t fix it”. To be able to adopt more modern computing-intensive applications like chatbots, they will need to change this mindset. 

 

That does not mean that the legacy systems have to be removed and replaced entirely. These can be connected to more modern applications through middleware interfaces and APIs.  

 

 3. Integrate with Internal Systems Relevant to Insurance Products  

 

In an ideal world, a chatbot will be able to interface with all the internal systems within a company. In insurance firms, the bot should at least be able to connect to some key systems in order to get the most value out of conversations with customers. For example, it should integrate with  

 

  • Document management tools – which helps insurers manage product and customer related documents effectively.   
  • Policy management software – which allows insurers to consolidate multiple policy systems and manage the full policy administration lifecycle for all lines of business.   
  • CRM systems – which is used to track customer interactions, log their sales activity and manage the sales pipeline  
  • Claims management software – which can save insurers resources by adapting to insurance software’s end to end capabilities for claims control   

 

4. Start Small for Omni-channel Integration    

Not all insurance providers may need to adopt a fully integrated omni-channel conversational AI system in one go. Depending on the number of channels you use to interact with customers, you can start automating the support in one of these. For example if your call centre query volume is primarily comprised of phone calls, you can start by exploring how many of these queries are repetitive that can be automated using a chatbot. Once you realise the productivity and efficiency benefits from this first step, then proceed to the email, messenger and physical mail channels. 

 

Starting small like this also helps you release the application faster and build on it over time. Scalability and the potential to iteratively improve is one of the benefits of AI applications, and companies can explore this to expand their use cases and capture increasing value over time. 

   

Back to contents




 

Future of Conversational AI in Insurance 

Semantic Search Will Help Demystify Product Complexity 

 

One specific way in which chatbots will dramatically change the insurance landscape is through semantic search. Semantic search is an informational retrieval technique where the engine looks for the intent and meaning behind a query to return a response. It goes beyond a simple lexical search where it looks for an exact match of the query words or its variants, without understanding the broader meaning of what is being asked.  

 

Using semantic search, it will be possible for bots to respond to queries whose answers are not directly or explicitly described in a chunk of text – and without any training. For example, it will be able to create an answer to a query such as “why is bread so fluffy” by referring a wikipedia article on the broad topic of bread.  

 

In an insurance context, this will allow bots to respond to queries about complex products by referring policy documents and product descriptions without being trained on those specific queries. 

 

Providers Will Go from Reactive to Proactive  

 

As conversational AI solutions become more sophisticated, we can expect the insurance industry to become less reactive and more proactive. With the proper data collection and integrations with consumers’ profiles, they will be able educate consumers on the insurance coverage corresponding to their shopping cart right when they make online purchases. 

 

We are already seeing major insurance providers integrating data from IoT devices and sensors to offer more personalised customer service. For example, AIA offers discounts for eligibly Vitality members on fitness programs and products using fitness trackers. Customers accumulate points for various fitness activities which can be exchanged for lifestyle rewards. They can also receive discounts on annual premiums, depending on their AIA Vitality status. 

 

Insurance Agents Will Work Alongside AI to Become Relationship Builders 

 

Policy holders and those looking to buy new insurance products have traditionally relied on agents for personal face-to-face consultations. Agents represent providers and are tasked with educating the consumer and answering all their questions before making the purchase. But with chatbots and conversational AI becoming more adept at automating most of the common repetitive queries today, there is a fear that that in the future, they will become sophisticated enough to completely replace human agents.   

 

Part of this is due to the robotic nature of conversational AI systems today. It has often been observed that users – both policy holders and agents – tend to bypass bots to try and chat directly with a live human agent, even if the bot is capable of providing factually accurate and helpful responses. This is where we could see a radical change in the future as conversational AI systems become more empathetic in their dialogue and users slowly get over their prejudices. 

 

The agents’ fears of being made obsolete can also be put to rest since even if some their traditional roles and tasks may become obsolete, they will still play a crucial role in the purchasing journey. This is due in part to the complexity of insurance products and the risk of making errors if consumers engage directly with the provider. Even today, consumers tend to prefer to engage with agents to make sure they understand all the details of their products, to make sure they are fully covered and that their claims are handled properly. The value of a skilled intermediary with a human touch cannot be underestimated, even after consumers learn as much as they can through chatbots. The role of agents in the future will therefore morph into that of product educators, process facilitators and relationship builders. 

Back to contents

 

Find out how KeyReply can help transform your insurance firm with conversational AI.

Request For Demo