Here's How an Employee-Facing Chatbot Helps You Manage Complexity

Tryg, Europe's largest non-life insurance company, uses an internal chatbot



Kindra Cooper
07/24/2019

chatbot

Chatbots are predominantly perceived as a customer-facing tool, but internal chatbots can be immensely helpful for training agents, managing complex back-office processes when used with robotic process automation, and facilitating information-sharing between teams through knowledge bases. 

Tryg, Europe’s largest non-life insurance company, uses an employee-facing chatbot to keep its sales representatives up-to-date on product changes across a portfolio of 38 different life, auto and home insurance products, each with their own distinct policies.

Most importantly, chat serves as a quick and easy interface for a complex knowledge base which sales reps can query while on the phone with a customer without putting them on hold.

The company hasn’t deployed a customer-facing chat function just yet, but Bjartmar Jensen, Tryg’s head of process excellence, says it was better to start with an internal chatbot so they could quietly test it, build up its knowledge base, and understand its use cases over a longer test period before rolling out an external version. 

Bjartmar Jensen Tryg“The internal chatbot is more demanding because it has to answer expert questions from employees,” Jensen told CCW Digital. “Questions from customers are less complex.” 

In his role, Jensen works to foster tighter collaboration between the sales and product development teams to achieve greater customer centricity. Given the sales team’s position on the frontlines with customers, they’re seen as a valuable source of voice of customer data.  Meanwhile, the product development team uses those insights as the basis for new product features and user experiences, and then communicates these updates back to the sales team via the chatbot. 

Read more: Innovation at Amazon - Why Good Writing Matters More Than Public Speaking

“With this knowledge pool, we ensure consistency in our outbound customer service and can program our brand experience into every conversation between customers and our employees,” said Eivin Joensen, an AI supervisor at Tryg who oversees the chatbot’s knowledge base. The chatbot from Boost.ai, known as Rosa, helps reduce the risk that a customer is greeted by an inexperienced support agent. 

A major mandate for the sales team is to build customer trust, insurance being a quality-of-life product category like healthcare where customers base their purchasing decisions on their confidence in the provider. 

Therefore, when a sales rep has all the right answers it generates credibility. Before Rosa, when a rep didn’t know the answer to a question, they would put the customer on hold, phone a member of the back-office team to consult with them and then resume the call with the customer. Now, they simply query one of their colleagues through chat. 

“We push messages every morning,” says Jensen. “So when the salesperson turns on their computer, it says, Good morning, Bjartmar. Remember today we launch this new product; you can read more information about it here.” 

Customer-facing chatbots are lauded for their ability to surface voice of customer insights including customer sentiment and intent based on the questions they ask and their responses. Similarly, Tryg uses data from its internal chatbot to determine if there are any gaps in its knowledge base and as a basis for training the AI. 

Read more: AI-Powered Knowledge Intelligence - The Next Gen of Knowledge Management

“We also use chatbot data from the internal chats to give feedback to the development department,” says Jensen. For instance, a high volume of questions about a given topic might reveal a product defect or incorrect information being disseminated. 

Customer-facing chatbots are rated on their rate of success in answering questions and understanding customer intent. Jensen says their internal chatbot is held to the same standards, where the team measures KPIs like intent recognition, missed triggers (rate of wrong answers) and from users giving specific responses a thumbs up or thumbs down. 

Currently, Jensen and his team are developing a robotic process automation capability for Rosa so that she can automate back-office tasks for sales reps, such as renewing a customer’s insurance policy. They’re also working to make Rosa more conversational rather than simply Q&A-based by training her to ask follow-up questions before unleashing her to customers. 

“For example, if a customer asks, How am I covered by my travel insurance? Then the chatbot asks, Are you going on vacation soon?” 

Doing so helps customers feel taken care of and listened to, while also surfacing opportunities for up-selling and cross-selling where relevant. 

RECOMMENDED