In 2017 Strayer University rolled out an innovative virtual machine-learning assistant called Irving, to help answer student questions online. Since then, Irving has handled over 740,000 conversations with students. Using Dialogflow, Google’s natural language processing tool, Irving can categorize student questions into “intents,” or topics, and then search campus databases through Google Compute Engine for customized answers. “We’re incredibly proud of it,” says Daniel McCarthy, SEI’s VP for Emerging Technology and Irving’s original lead developer. Surveys show that students have responded favorably as well: as of January 2020, 89% of Strayer students agree that Irving helped them solve their issue easily and 93% of them would recommend Irving to other students.
Joe Schaefer, Chief Transformation Officer at SEI, was pleased with how smoothly the implementation went: “we originally thought we’d need on-site consulting, but we found that with our team’s creativity and desire to learn, plus Google’s tools and people to point us in the right direction, we could build a great product for our students on our own.”
By using Google technology to create Irving we’ve actually seen demand for support increase – in a good way. Because Irving is so helpful, students are asking him more questions—and getting more answersJoe Schaefer, SEI’s Chief Transformation Officer, Strategic Education, Inc.
Saving time and resources with machine learning
Strayer University’s students are older than traditional college students and are often taking care of family and working in addition to attending classes online. “Busy working adults find tremendous value in 24/7 support,” Schaefer says. “There’s a customer experience challenge we’re trying to solve. Our students don’t have to waste time calling us, Irving can now answer the majority of their concerns. Many of our students are back at school after a long time away and may need extra support.” One student on a survey noted that “I like that Irving is available 24/7 so that if I have a question, I can ask it at any time. Also, if I have a question where I may feel dumb for asking a real person, I am not as embarrassed to ask an automated system.” Other respondents mentioned that using Irving was faster than a phone call and if they had repeat questions “you don't feel like you keep bothering a real person.”
By shifting phone calls to automated chats and offloading repetitive tasks, Irving also frees up Strayer staff to serve students with more complicated questions and needs. This improves service as well as administrative efficiency across departments, with Irving trained on about 1,500 different topics so far. The team had transcripts of student questions to work from in assessing where Irving could make the most impact. Schaefer says, “degree program requirements is a good example of how Irving can be proactive because there are many requirements people don’t even know to ask about. If a student asks, ‘what classes do I have remaining?’ Irving can check courses taken and passed, suggest courses for the future and confirm current enrollment status for the current and upcoming quarter. Also, Irving has capabilities across domains and might know more than a person answering the phone.”
Getting more answers and solving more problems, faster
McCarthy points out that Irving’s call logs reveal patterns of usage or even pain points that wouldn’t otherwise be visible: “if we notice a cluster of problems with registration we can use that information to improve our business processes. We’ve already reduced the time staff spent on the phone and we can have more predictable schedules for staffing.” Irving was trained to try to answer any incoming question three times before referring the student to a human. McCarthy reports that, as of January 2020, only about seven percent of the conversations have had to be transferred to staff members. Schaefer adds that Irving increases demand as well as capacity, so now students are simply asking more questions—and getting more answers.
So far, the team has focused on providing the best support possible for students and employees, but they believe this technology can directly help learning too. “We want to make learning truly digital, with immediate feedback loops to maximize students’ potential,” McCarthy says. “By partnering with Google, content providers, our professors, and others, we can help students develop meta-cognitive skills to figure out for themselves what they know. They can become engaged learners.” “Artificial intelligence is going to fundamentally reshape how we interact with everything, including school. We’re still at the start of this research and development journey,” Schaefer concludes. “Like Irving, we’re building our expertise.”
I like that Irving is available 24/7 so that if I have a question, I can ask it at any time. Also, if I have a question where I may feel dumb for asking a real person, I am not as embarrassed to ask an automated system.Strayer University student