Suddenly, over just a single weekend, they had to shift the global workforce of over 22,000 people to working remotely.
Not surprisingly, existing processes and workflows weren’t equipped for this abrupt change. Customers, employees, and partners — many also working at home — couldn’t wait days to receive answers to urgent questions.
As we began building the framework and interfaces on Slack, they realized the same, specific questions and issues were coming up frequently. By focusing on the most common and weighty issues, they decided to optimize support for frequently asked questions and issues. They dubbed this AI and machine-learning-based Slack channel “#wfh-support,” and it had built-in natural language processing (NLP).
The chatbot’s answers could be as simple as directing employees to an existing knowledge base article or FAQ, or walking them through steps to solve a problem, such as setting up a virtual private network.
Clear results – satisfied employees
The results have been remarkable. Since the initiative went live on April 14, the automated system has responded to more than 3,000 queries, and they have witnessed significant improvements in critical areas. For example, they noticed more employees were seeking IT support through email when we shifted to work from home, and it became important to decrease the turnaround time on email help tickets. With the help of a deep learning and NLP based routing mechanism, 38% of email tickets are now automatically routed to the correct support queue within six minutes. The AI routing bot uses a neural network-based classification technique to sort email tickets into classes, or support queues. Based on the predicted classification, the ticket is automatically assigned to the correct support queue.
This AI enhancements has reduced the average time required to dispatch and route email tickets from about 10 hours to less than 20 minutes. Continuous supervised training on the routing bot has helped them reach approximately 97% accuracy — nearly on par with a human expert. As a result, call volumes for internal support have dropped by 35%.