Total 640 words used.
After analyzing our support data in 2021, we identified significant issues with long waiting times for basic and advisory questions raised by iERP users (iERP - integrated Enterprise Resource Planning - SAP based platform used by Quote to Cash teams for processing orders). This has placed a burden on iERP EMEA Super User team, as they have had to handle many simple questions and could not focus on more complex issues.
To address these challenges, we proposed implementing chatbot technology, which would automate and eliminate the need for manual, repeatable efforts in responding to user inquiries. By doing so, we can greatly improve the speed at which we address these inquiries.
Approach and solution:
We worked throughout 2021 to create a chatbot by exploring various available technologies and building a management system to provide the chatbot with a corpus of over 2000 questions covering different areas of support and brands within the scope of the EMEA Super User team. The initial version of the Agent Assist (AA) chatbot was deployed in December 2021. The Agent Assist chatbot uses a cloud-based solution such as Red Hat, OpenShift, Kubernetes Services, and IBM Cloud, and incorporates AI technologies like Generic Propriety NLC (Natural Language Classifier) and ELT (Elastic Training), an AI machine learning model, and algorithm.
Right from the moment of its deployment, the solution gained significant traction with its users. Our project achieved a significant improvement in the efficiency of our customer support operations. By the end of 2022, 90% of advisory questions that could be addressed by our Agent Assist application were being handled through that channel, with only 10% still being handled through our legacy Swifty system. This shift resulted in faster response times and increased customer satisfaction, as our customers were able to quickly get the assistance, they needed through our more effective Agent Assist application.
In parallel, the chatbot was improving thanks to its learning corpus curated by the Super User community, enabling the solution to better understand inquiries and expanding its knowledge base to cover a wider range of topics.
Innovative features cutting edge in digitalization:
The Agent Assist chatbot, which is accessible via Slack or a web application, leverages the natural language and learning capabilities of Watson Services to answer user questions and provide quick support for simple issues.
The project team has recently migrated to a new platform to take advantage of a wide range of advanced features, including Tuneable Auto Curation, which enables the team to configure the auto curation process with multiple passes to produce more accurate results. It is timesaving and improves the quality of the information presented to our customers.
- us to gather and analyze a vast amount of data from the web. This feature enables our chatbot to provide more accurate and relevant answers to customers' inquiries.
- continuously train and improve the performance of our chatbot, stay up to date with the latest trends and provide optimal assistance to customers.
Generic Proprietor Training model is a new answering system that integrates a proprietary AI-based training model with specific machine learning algorithms. It provides predictive and suggestive functionality of the chatbot to improve overall efficiency.
Impact:
Thanks to the deployment and continuous enhancement of Agent Assist, our teams were able to save over 10,000 hours of waiting time in their business support cycle in 2022. This project enabled our professionals to speed up the support for transactions from IBM clients by one day, which was the standard waiting time for simple advisory queries before the deployment of Agent Assist. This has empowered the teams to be more self-sufficient when working with iERP, and at the same time, helped Agent Assist to continuously improve. Additionally, EMEA Super Users were able to focus their time on more complex tickets with faster resolution, thanks to the 23% reduction in Advisory tickets received year-on-year.