The hype around Artificial intelligence is in full swing. Its cheerleaders claim it will rapidly transform sector after sector, from banking and insurance to retailing and the auto industry.
It is capable of plugging into many areas of the insurance process, from powering underwriting decisions with machine learning to speeding up claims assessment processes. Insurance is an industry built on data, so it should provide fertile soil for the seeds of AI.
Yet it is far from clear that the conservative insurance business, long steeped in traditional ways of operating, will be disrupted by technology at the same speed as sectors such as banking, retail or entertainment.
Smart Data Analysis
Insurance giants are already experimenting with AI as new entrants develop the technology. US insurgent start-up Lemonade offers home cover policies in just 90 seconds using AI and has reduced the time required to process claims payouts to three minutes by applying smart data analysis. Switzerland’s Zurich Insurance is using AI to decide on personal injury claims after tests cut the processing time from an hour to a matter of seconds.
The venture arm of Australia’s largest insurer, QBE, recently acquired start-up HyperScience, which uses machine learning to automate office processes.
It has also put money into RiskGenius, which integrates AI to analyze policies and create new products, and Cytora, which uses risk-scoring based on AI to make underwriting decisions. The start-ups are looking to sell their services to the giants, rather than seeking to compete with them head-on.
Handling Customer Enquiries
In addition to insurtech start-ups, two rival AI systems are vying for the business of insurance giants. IBM Watson and IPsoft’s digital employee known as Amelia AI are both the subject of talks with major banks and insurance companies. This week, IPsoft announced that Spanish bank BBVA will start using Amelia.
IPsoft’s AI system is already in use at Allstate Insurance in the US. The insurer’s 350 call centre employees have been using Amelia to solve customer enquiries, putting queries to the chatbot for fast and efficient answers.
Allstate says Amelia has been involved in three million conversations since it was introduced last September, helping to cut the average call duration by nearly 10% to 4 minutes and 12 seconds.
Natural Language Processing
IPsoft makes big claims for Amelia, saying that its level of human interaction makes it suitable for both back- and front-office use. Companies can use Amelia as an assistant for customer agents or as a chatbot in direct communications with customers.
Many chatbots simply act as search engines, finding the answers to frequently asked questions from Excel spreadsheets. They categorize queries into a series of buckets and are measured by the number of buckets they use.
IPsoft says Amelia is different because it uses natural language processing and employs ways of mimicking the human brain to converse in human-like ways. The system is divided into six “sub-brains,†each with a different function, to help it mimic thought and conversational processes.
Emotional Memory
A “semantic†sub-brain reads FAQs from Excel, while the “logic†portion of the brain makes inferences from what people say. The third level is the “social talk†brain, which enables Amelia to chat casually about subjects such as the weather or sports, while an “emotional memory†brain interprets the sentiment and mood of a speaker and tries to build that knowledge into the conversation.
The “process memory brain†enables Amelia to follow standard processes, such as the steps to buying an insurance policy. The sixth brain is “episodic memory,†which learns from real life conversations between humans to answer questions.
All six sub-brains work together to answer a query. If for instance, a customer rang in to report an injury, Amelia’s emotional brain would prompt it to say: “I’m sorry to hear that.â€
Fraud DetectionIn addition to Allstate, IPsoft, founded by Indian-born mathematician Chetan Dube, is working with insurers including MetLife, AIG and Liberty Mutual to implement Amelia in their operations.
AI has great potential in other areas of the insurance business. Machine learning can improve fraud detection by identifying typical patterns in fraudulent claims and flagging them up to enable an agent to ask clarifying questions. It could also render underwriting more effective, helping insurers make their questionnaires shorter and more relevant.
As a customer service chatbot, Amelia can handle complex queries, not just straightforward processes such as: “I want to make a claim.†But importantly, it identifies the point at which a customer’s query needs to be referred to a human agent.
Compliance Requirements
The extent to which AI will change the insurance industry, and the speed at which it might do so, depend on a number of factors. A big issue will be the effectiveness of the technology in improving the claims ratio by eliminating fraud and making the underwriting process more efficient.
The heavily-regulated insurance industry is protected in part by its complex bureaucracy and compliance requirements, alongside significant barriers to entry such as the need for risk assessment expertise. AI will undoubtedly bring change, but it is likely to be a gradual evolution rather than a sudden transformation.