Revolutionizing customer service

By | 5 minute read | October 22, 2020

People are drowning in the oceans of information and data volumes are increasing exponentially—making it harder than ever to find the right information at the right time. We know now that AI can be a life raft, providing new ways to interpret what people are looking for and finding it quickly. By using IBM Watson for its AI-powered search and retrieval solution, Enzyme Advising Group helped simplify a time-consuming process for Ibercaja Bank. The solution made Enzyme Advising Group a 2020 IBM Beacon Award winner for Outstanding Data and AI Solution.

Trawling for data
Ibercaja Bank
had a data challenge to solve. As a broker for Spanish insurance company Caser, it arranged insurance contracts with customers and then helped them to file claims. However, legacy search technology and claims process were slowing the process down.

Customers filing a claim must visit an Ibercaja branch and talk to a bank teller, but those employees are not insurance experts. Insurance claims can get complicated and clerks must follow different processes, depending on the nature of the claim. Tellers had to search for the correct process document before they could help a customer. Harald Messemer, Enzyme Advising Group’s COO, explains:

“The bank had a transaction-based procedure management system to manage those written processes” and he adds that there were between 80,000 and 90,000 banking procedure documents in the system. Finding the right instructions for an insurance claim was an exercise in frustration.

The search method was counter-intuitive because it relied on document names. Tellers handling complex or unusual claims often wouldn’t know what to search for, prompting them to call the contact center while the customer waited. Each month, Ibercaja employees receive nearly 30,000 general support tickets per month, 2,000 of which comprise insurance claim tickets that the chatbot could help streamline.

“If the bank clerk doesn’t know about a process and has to call the contact center to know how to do it, the customer ends up waiting for another five or ten minutes,” Messemer says. “If the employee cannot get a timely answer, some customers will end up abandoning the process.”

The bank wanted to cut customer waiting times and reduce the time that contact center agents spent handling queries from tellers. These agents are highly skilled, and the bank wanted to put their time to better use.

Chatbots to the rescue
Ibercaja envisioned an AI-powered solution that could interact with agents in more intuitive ways. It created a self-service chatbot system that enabled 4,000 employees at the bank to query back-end systems for process documents using natural language, according to Messemer.

Enzyme Advising Group worked with the bank’s business staff in an agile design and development process, creating a workable application in ten weeks. It used the SCRUM agile development process for weekly sprints, creating more refined iterations of the system over time. This enabled the firm to evolve the system iteratively as it learned more about the banking employees’ requirements.

The Ibercaja team had no experience in designing a chatbot, so Enzyme began by creating a simple prototype natural language query system in the first week of development to demonstrate what was possible. It used a series of questions and answers to understand the nature of a claim and narrow down the possible list of documents. “We’ve been waiting for something like this for eight years!” exclaimed one staff member at the bank.

From there, the development team worked with the bank’s business staff on a fully functional chatbot system. Using its consulting experience, Enzyme learned from business employees how to find the appropriate documents in the bank’s vast collection.

The final system consisted of two components. The bank wanted to host Orchestrator, a JavaScript-based user interface for the system, in an Azure cloud environment. This communicated with the Watson AI system in the IBM cloud and used two APIs to deliver the chatbot service: the Assistant and Discovery APIs.

Together, these components enable the service to integrate with a variety of front-end applications, including not just web sites but also mobile apps and messaging platforms such as WhatsApp. It can also integrate with enterprise applications such as customer relationship management and robotic process automation systems.

A powerful relationship
IBM was a crucial partner for Enzyme during the development and deployment stage. The consulting company has worked with IBM Software for four years and has a strong relationship with its team in Spain.

IBM helped Enzyme to overcome any speed bumps, explains Messemer. These included commercial issues like pricing out elements of the system that the client wanted to be deployed on Azure rather than on IBM’s own cloud.

IBM was also able to use its existing relationship with the bank to help push the deployment of the system in six months. “It was a question of relationship management, helping the client to understand technology they hadn’t used before, and to put all the pieces together,” Messemer says.

Measurable benefits
Having deployed the chatbot system, Ibercaja was able to reap the rewards. The solution passed user acceptance testing on the first try, which is a testament to the enhanced user experience inherent in conversational systems. It has processed 1,900 documents with a 75% probability of finding the right document in the top five results, Messemer said. Four times out of 10, the chatbot’s first suggestion is the correct document.

The chatbot system complements its high accuracy with fast operation. It delivers its results in a maximum of five seconds, meaning that customers don’t have to wait for answers.

Since then, Enzyme has taken the system still further. It set up a monitoring system for Remedy, the helpdesk system that the bank’s contact center uses for filing claims. The monitoring system watches Caser’s back-end claims processing software to monitor claims status, enabling bank tellers to know how a customer’s case has progressed. When a customer comes in to check on their insurance claim, the bank teller can give them an instant update.

Enzyme Advising Group has turned its Watson-based chatbot solution into a conversational search product that generates repeatable revenue. It has already helped at least ten organizations across sectors including e-commerce and education, picking up entirely new clients in the process. Thanks to a powerful combination of commercial partnership and cutting-edge technology, this is a success that will live on as a marketable solution.

Beacon Award winners are changing the world through their solutions. Learn more today.