As people increasingly turn to online channels for insurance information and claims processing, OP wanted to ensure it could give its customers helpful guidance without delay.
OP worked with IBM to create an intelligent chatbot, training IBM® Watson® Assistant to understand common insurance issues and learn Finnish — enabling it to interact with customers in natural language.
Improves serviceas customers don’t need to wait for human agents to become available
Automates twice as many conversationsas expected, reducing the need for human agent intervention
Accelerates ROIby scaling customer service capabilities without additional headcount
Business challenge story
Scaling a successful service
OP is one of Finland’s largest and most innovative financial services companies, offering not only traditional banking, investment and insurance products, but also diversified services in other sectors such as healthcare and mobility.
The company is constantly looking for new ways to improve customer service and increase operational efficiency, and has identified data science and cognitive solutions as key tools in developing new services.
Jaakko Sävilammi, Development Manager in the company’s data science unit, explains: “Our role is to explore how cognitive solutions can help OP solve real-world business problems, so when the insurance division approached us about finding a way to scale their customer service center, we were eager to help.”
To proactively meet growing customer demand for online services, OP’s insurance division launched a now-popular website that gives customers insurance advice. A key part of the website is an online chat feature, which customers can use to interact with customer service teams and ask questions about their claims.
“When we launched the chat feature, it was an instant success, and the number of users grew by 1,000 percent in the first few months,” says Sävilammi. “The problem was that we needed a customer service agent to respond to each conversation — and as the popularity of the service continued to grow, we were concerned that the customer service team would become overloaded.”
OP realized that customers are most likely to seek insurance advice when they need to make a claim, for example, after a car accident, a healthcare diagnosis, a flight cancellation or a burglary. At this critical time, the last thing they want is to have to wait for 10 minutes for someone to reply to their chat message. OP wanted to ensure that it could provide rapid service to its customers and help them get their lives back on track as quickly as possible.
“The classic problem for customer service teams in the insurance industry is that when there’s a major problem, like a severe winter storm that makes driving dangerous and forces airports to close — that’s when the most customers need our help,” says Sävilammi. “But because staff availability is limited, it’s also the time when customer service is slowest and waiting times are longest.
“We realized that if we could harness artificial intelligence to help our customer service teams handle these peaks in demand, we could make a significant impact on customer experience.”
Putting AI to work
OP’s data science team began researching whether an AI-powered chatbot could help to handle some of the incoming requests from the online chat service.
One of the main challenges was that the chatbot would need to be trained not only to give insurance advice, but also to converse with Finnish customers in a natural way. The company needed a vendor that could not only provide the best technology, but also an implementation team who understood the Finnish language and culture.
“We tested solutions from many vendors, and IBM Watson Assistant produced the best results,” says Sävilammi. “We were impressed by IBM’s AI heritage with Watson™, and with their roadmap and investment in the field. Moreover, IBM has a large presence in Finland and was able to assign local resources to the project. Having native Finnish speakers on the team is a huge advantage for this kind of natural language processing project.”
OP’s initial research suggested that training the chatbot would require a significant time-investment, and that it would take many iterations until the AI model achieved the desired level of accuracy. However, by working with a team from IBM Services™, the company was able to take full advantage of the Watson Assistant platform and deliver a solution on time and within budget.
“Of course, like any project, there were challenges,” says Sävilammi. “But IBM helped us take a very pragmatic approach. We used the early iterations of the chatbot as an internal tool that our customer service teams could ask for information — so it started delivering value quickly. This also helped us continuously assess the accuracy of the model in a real-world context, so by the time we were ready to roll the chatbot out to our customers, we were already very confident that it would provide a good user experience.”
One of the key pieces of advice that OP would give to other companies is that AI projects require a diverse set of skills to succeed. In addition to data scientists with expertise in natural language processing, the OP and IBM team included business people who understood the insurance domain and the data, IT people who assisted with application development and integration and customer experience experts who played a vital role in fine-tuning the chatbot’s interactions to ensure that users would stay engaged.
Sävilammi adds: “Teamwork is vital because supervised learning is a long, resource-intensive process. You need people who have the motivation, expertise and patience to power through the many iterations of design, training and evaluation. IBM made a critical contribution, augmenting our in-house resources and helping us build up our own expertise.”
At the start of the project, OP established a set of metrics that it would use to evaluate the chatbot’s performance.
“Our main objective was that the chatbot should be able to handle a certain percentage of conversations with customers from end to end, without human involvement,” says Sävilammi. “When we rolled out the chatbot to our customers, it outperformed our targets by a huge margin: the rate of automation was more than double our expectation.”
OP customers don’t need to wait for a response to their chat messages, as chatbot and customer service teams work contiguously. This improves the customer experience, especially if the user is in a stressful situation and needs an answer quickly.
“IBM Watson Assistant enables us to scale the number of customers we can support, without having to increase the size of our customer service team,” says Sävilammi. “As demand for the online chat service is growing all the time, we expect to see a strong return on investment.”
He concludes: “It’s rare for a vendor to keep all their promises, but IBM delivered every iteration on schedule and on budget, and the results were twice as good as we expected. The whole project has been a great example of what can be achieved when software providers are prepared to listen and innovate, rather than just pushing their products. Working with IBM has been a breath of fresh air.”
OP Financial Group
Established in 1902, OP is one of Finland’s largest financial services companies. It employs around 12,000 people in Finland and the Baltic countries, and operates in a diverse range of markets, including banking, life and non-life insurance and wealth management, as well as branching out into health services, with five hospitals across the country.
Take the next step
To learn more about the IBM solutions featured in this story, please contact your IBM representative or IBM Business Partner, or visit the following websites:
IBM is working with organizations across the financial services industry to use IBM Cloud™, Cognitive, Big Data, RegTech and blockchain technology to address their business challenges. Watson Financial Services merges the cognitive capabilities of Watson and the expertise of Promontory Financial Group to help risk and compliance professionals make better informed decisions to manage risk and compliance processes. These processes range from regulatory change management to specific compliance processes, such as anti-money laundering, know your customer, conduct surveillance and stress testing.