To NLP or not to NLP… wait, what’s the question?
What’s the question, indeed. Take a look at the simple question below with four simple words. Notice that depending on what emphasis you place on which word, the question has a totally different meaning:
Did you buy that? Translated meaning: Or did you borrow it?
Did you buy that? Translated meaning: Did someone else buy that?
Did you buy that? Translated meaning: Did you steal that?
Did you buy that? Translated meaning: Wow, that’s ugly.
Welcome to the English language, or any language for that matter. It’s filled with nuances, syntax, homonyms, synonyms, and even slang. It’s no wonder that when a computer can respond correctly to a typed or spoken sentence it’s super cool! This is Natural Language Processing (NLP) at its best.
In 2017 and 2018 BetterBot built the largest proprietary NLP engine in all of Multifamily filled with the terms we use every day in our Industry. Then walked away from it altogether. Why?
Because NLP breaks 3 out of 10 times.* It turns out the biggest and best NLP solutions in the world such as IBM Watson, Google Dialogflow, Amazon’s Alexa or Comprehend, etc., all have the same breakage problems.* And these are the best on the planet!
Imagine asking a leasing agent the same question ten times and three of those times, you didn’t get the answer you were looking for, or worse, you didn’t even get a response back. That can profoundly affect your user experience. With this type of error rate, NLP requires a level of human supervision to course-correct when it incorrectly interprets an interaction. This doesn’t make leasing jobs easier, it actually creates more work for leasing teams now having to either monitor, assist or completely take over when NLP fails.
Taking a look at some of the biggest conversational bot solutions such as Drift and Intercom, they have all moved to Guided Conversation or Rules-Based Chat. The reason being? You can’t break it. If you look at some of the other bots in the Multifamily space, they recently began incorporating some sort of Guided Conversation because they know that without it, the system can’t stand alone.
Guided conversation can also lead renters through different levels of a single inquiry, providing deeper understanding into the topic and thus making them a more qualified prospect when they ultimately convert to a tour. Natural Language Processing assumes the renter wants to continue the conversation after their question has been answered, rather than presenting additional routes to follow.
BetterBot learned a great deal during their early development period about what prospective renters and residents are looking for in their digital interactions. Approximately 40% of consumers don’t care whether a chatbot or a real human helps them, as long as they’re getting the help they need (Invesp). So, let a chatbot be a chatbot and own it.
And finally, IBM found that Chatbots can help reduce customer service costs by 30%, services such as answering Frequently Asked Questions, scheduling tours, and providing renter resources. Not to mention, always being available, 24 hours a day, 7 days a week, 365 days a year. We even throw in an extra day every four years, just for good measure. So let’s be clear, Chatbots are NEVER going to replace a leasing agent, but it can help offload these menial tasks so that leasing agents can go back to doing what they do best… being human!
*BetterBot 2017 & 2018, LivePerson 2016
