Questions for VBZ? The AI Butler Assists Customer Service with Answers.
Contact's customer service. To streamline response times and ease employee workloads, the company tasked UMB with developing a proof of concept (PoC). The goal: use AI to handle 20% of rate-related emails. The solution leverages watsonx and IBM Cloud Services.
#IBM #Artificial IntelligenceThe Zurich Transport Authority (VBZ) operates 14 tram lines, 6 trolleybus lines and 53 bus lines, covering a total distance of approximately 294 kilometers. The VBZ fleet includes modern, environmentally friendly vehicles, many of which are low-floor for barrier-free access. Ticket prices and zones are integrated into the ZVV (Zürcher Verkehrsverbund) system, allowing easy transfers between different modes of transport.
Lots of customers, numerous contacts
Supporting a large number of customers requires efficient service. At VBZ, the high demand was challenging the customer service team. Every day, many emails with different requests had to be processed manually. The average response time was ten minutes, while more complex requests could take up to 30 minutes. This often resulted in wait times and backlogs. To alleviate this problem and increase efficiency, UMB AG, together with VBZ and IBM, developed the «VBZ E-Mail Butler» as a proof of concept (PoC). A proof of concept provides evidence that an idea or solution works in practice. It is usually carried out in the early stages of a project to check feasibility before larger investments are made and helps to identify potential problems.
People have the last word
UMB uses cutting-edge technologies such as analytics, machine learning, and automation to optimize processes across its broad industry portfolio. In this project, UMB contributed key AI and software expertise and was responsible for the technical implementation of the PoC. The Email Butler PoC aims to automate email processing at VBZ by having the butler generate responses to a variety of queries and categorize incoming emails. The emails are not answered directly by the bot - instead, employees address the queries to the chatbot, which provides a suggested response that is checked, corrected if necessary, and then sent. People remain the final authority.
IBM's watsonx underpins this AI solution
The VBZ Email Butler works with watsonx.ai, IBM's AI platform for enterprises. Watsonx.ai uses machine learning, natural language processing and data analytics to deliver scalable solutions that integrate seamlessly with a variety of AI tools. The platform makes it possible to quickly implement AI applications in the enterprise and solve even complex challenges. In this case, source information from VBZ, ZVV and SBB and, above all, previously answered questions form the basis for the answers. The goal is to ensure that the answers provided by the e-mail butler match the tone and content of the VBZ. UMB's AI experts ensured the successful and seamless implementation of AI into VBZ's existing processes.
Intensive tests of various AI models
VBZ, UMB and IBM worked closely together to ensure that the solution was precisely tailored to VBZ's operational requirements. The project began with a detailed analysis of the existing processes, during which the team - consisting of UMB AI experts and four IBM employees - examined the email workflows, identified key problem areas and defined the main objectives for the PoC. In an intensive test phase, various AI models were tested for their quality of response and adapted to the feedback provided by the VBZ employees. The first promising results of the PoC soon became apparent: VBZ E-Mail Butler was able to generate high-quality response suggestions for around 20 percent of queries, significantly reducing the processing time for standard queries.
ZVV-Contact customer service will have more time for customers
The most important result of the PoC is significant time savings. Automated responses to recurring requests allow agents to focus on more complex and urgent cases.«The VBZ email butler has shown that a bot can answer simple queries with high quality. Implementing this solution will free up a lot of time that we can use for more complex requests», says Benyamin Svensson, team leader at ZVV-Contact.
Following the successful PoC, VBZ is now looking to extend the AI capabilities to other communication channels and respond in near real-time during peak times.
Our AI & Data Scientist - turns your data into_valuable knowledge.
The VBZ email butler has shown that a bot can answer simple queries with high quality. If we implement this solution, it will free up a lot of time that we can use for more complex requests.
