In his free time, Jeremy Grafke is a mild-mannered Kansas Citian working toward his MBA. But when employees or customers of LeaderOne Financial Corporation need help, he dives into action, answering their questions and leading the support automation initiative.
Jeremy is a flesh-and-blood L.O.I.S. Copilot at LeaderOne, a national mortgage lender. L.O.I.S. is LeaderOne’s chatbot, powered by Capacity. Short for “LeaderOne Information System,” L.O.I.S. is a tireless, always-on digital worker who answers users’ questions through a friendly AI-powered chat interface.
But L.O.I.S. can’t answer everything. Unlike humans, computers can’t receive information from multiple sources at the same time, break it down, and apply logic to it. Countless subtleties such as context, tone of voice, and past experiences also play a role in their decision-making. Even computers can’t process all of those inputs like humans can, but with a little help, they can acquire more information easily.
That’s where Copilots like Jeremy come in. They are the humans-in-the-loop who complete the cycle of knowledge sharing for LeaderOne and other companies using Capacity’s support automation platform.
“When L.O.I.S. is unable to answer a question, either from within the company or outside the company, she generates a ticket that comes to me,” he explains. “I answer them if I’m capable, and that information goes back into the system. But because I’m not a mortgage expert, LeaderOne has added representatives from every department who I can refer questions to. Then, that information goes into L.O.I.S. so she can answer the question the next time.
“I’m like a quarterback, and the department heads are my receivers for answering customer and employee questions. You can think of L.O.I.S. like Siri or Alexa. I help build her out and teach her new things.”-Jeremy Grafke, LeaderOne
Providing great customer experiences.
Buying a house is stressful. Borrowers have questions at every step of the process, from how to qualify for a loan to what to expect at closing.
Fortunately, many of these questions have standard answers that a solution like L.O.I.S. can manage. By crawling LeaderOne’s dynamic knowledge base of loan documents, state and federal regulations, and more, L.O.I.S. can usually find solutions.
But if she is unable to find an answer, or misunderstands the question, Copilots like Jeremy or the department heads can help. After a CoPilot or department head responds to an unanswered inquiry, Capacity automatically and instantly shares the knowledge with the customer, and the answer is added to the knowledge base. The cycle of knowledge sharing is complete.
“I’ve worked a lot of public-facing jobs before, so I understand the importance of great customer service,” Jeremy says. “I enjoy answering customers’ questions when I can, but when I am unable to, the questions are mapped to an appropriate response, or new exchanges are created by mortgage experts and added to the knowledge base.
The result? A smarter knowledge base capable of answering more questions, giving Jeremy time to perform other tasks—like assisting other LeaderOne employees.
Delivering better employee experiences.
Even mortgage experts need access to L.O.I.S.’s dynamic knowledge base sometimes—or, they may have questions about less technical things, such as dress code or PTO.
L.O.I.S. and Jeremy are there for them, too. But LeaderOne employees have even more options for getting help.
“Employees can access L.O.I.S. either through the web concierge or through whatever collaboration tool they are using, like Microsoft Teams or Slack,” he explains. “It’s very easy for them to reach her and get their questions answered just like they are talking to a coworker.”
And if L.O.I.S. doesn’t know the answer? Again, it’s Jeremy to the rescue.
“I know immediately whether the question is coming from a customer or an employee,” Jeremy says. “That allows me to tailor my response appropriately and gives me clues into where to direct the question, whether it’s HR or IT. And once the question is answered, the question and response feeds back into L.O.I.S.’s knowledge base.”
Making copilot’s lives easier, too.
If you are thinking Jeremy must be a technical wizard to do all of this routing and re-routing of customer and employee questions, he has a secret for you—he has a degree in Sports Management from Wichita State University in Kansas, but no formal IT training.
“L.O.I.S. is very user-friendly. The back-end is well laid out and easy to navigate, and the front-end is simple to interact with.”
Essentially, managing an AI interface like L.O.I.S. is like managing an inbox. She organizes groups of relevant changes intuitively, matching them to the correct conversations. This allows her to learn not only the answers to the questions, but how similar groups of questions relate to each other.
“I think it’s a convenient tool, for sure,” Jeremy adds. “It’s definitely a big convenience for us, because we get instant feedback. Instead of emailing a department with a question and it taking hours or days to get an answer, I can usually get it back in a matter of seconds. The speed in getting what you need, whether it’s me, a customer, or an employee, is one of the biggest values.”
Jeremy says it’s also easy to add new information to L.O.I.S. For example, in the mortgage industry, new regulations may need to be fed into her knowledge base.
“We have L.O.I.S. connected to several of our key systems, including the Encompass Loan Origination System,” he explains. “It catches most of those things automatically, but if something needs to be entered manually, it’s pretty simple. It’s really just a few clicks. Everyone that I’ve trained on LOIS has been able to catch on really well.”
Imagining the future with L.O.I.S.
Humans-in-the-Loop like Jeremy do what AI systems like L.O.I.S. can’t. They interpret nonverbal and other cues and listen for nuances that computers can’t detect—yet.
However, they are getting smarter all the time. With advances in natural language processing, the handoff to a human-in-the-loop is becoming smoother, the ability to understand questions in multiple languages is improving, and the power to detect human sentiment is growing.
The last advancement may be the most important. For example, think of the different emotions associated with a single word, like “right.” There’s a big difference between a customer who says “Right!”, “Right?”, and the sarcastic “Riiiiiiiiiiight …” Correctly interpreting emotion is key to identifying and resolving questions efficiently, for both customers and employees.
After all, for Copilots like Jeremy, the most important thing is getting the right answer to the user.
“I’m not sure that the average customer has a preference between communicating with a human or a chatbot like L.O.I.S., as they get their question answered quickly,” he says. “I feel that folks, especially nowadays, after pretty much the whole world has been virtual for over a year, have higher expectations for automated tools like L.O.I.S., expecting them to know everything—but they are still understanding when they don’t.”
Jeremy finds working with L.O.I.S. rewarding, as well. “I’m basically building out our own AI system,” he says. “It’s like being the heart and soul of LeaderOne’s knowledge base, and I enjoy working with the team of subject matter experts here. Everyone here loves it.”