Chatbots have been around for some time. Some say that the first one was developed by an MIT professor in 1994. We use them every day as consumers, often without even knowing it, with applications like Facebook messenger, WhatsApp, Alexa, Google Assistant, and others. We enter text into an app, or we speak into an instrument, asking questions, and the chatbot responds with answers.
Many people think chatbots are AI. The fact is that chatbots do use AI, but they do so in order to understand the text or the language that they’re trying to consume. There is a whole field of computer science called Natural Language Processing (NLP) involving linguistics, information engineering and AI that is designed to improve computers’ understanding of human language.
But chatbots are not AI. Chatbots use NLP to understand the text or the language and then they choose a fixed step or steps to perform. This can come from programmed content (“What are your store hours?”) or it can come from interfacing with backend systems (“When will my shipment arrive?”). The answer chosen by the chatbot is determined solely based on NLP.
As the B2B world begins to look at AI, particularly for service and support, chatbots are a natural first place to look. The possibility to deliver new capability through a tool already used by many lowers the bar to adoption. In addition, NLP has markedly improved over the past few years, which makes it seem more attractive as a solution. And it is a good solution, particularly for simple or frequently recurring tasks.
So what are the challenges with chatbots?
The biggest challenge is that everything a chatbot does must be programmed, just as if it were an application. Questions and their answers must be written in code. The code needs to be compiled and then run through QA. If there are changes, the code must be re-written, de-bugged, and run through QA again. If the original authors of the code are no longer available, this presents further challenges.
Other potential issues:
- The more complex the task, the more programming must be built and maintained;
- Chatbot uses AI to understand language then chooses a fixed step to perform. The step chosen is determined only based on the language provided;
- A Chatbot does not learn based on the success or failure of its answers;
- Despite the advancements in NLP, there remain challenges with speech recognition;
- Different humans may describe the same request in different languages.
So how is RevTwo AI not a chatbot?
RevTwo AI is typically used by field service, product support or even end-users to rapidly triage and resolve problems with complex machines in the field. It is a system that diagnoses problems from symptoms. Usually, a machine has some indication of what’s wrong, and this can be identified through electronic data such as error codes, readings, misconfiguration, and other symptoms. There are also visual, tactile or audio indications such as leaking fluid, smoke, or sounds. RevTwo consumes these symptoms and observations and uses AI to propose a solution or guidance towards a solution.
Each new diagnosis trains the AI and makes it smarter. As more and more users utilize the AI, it eventually becomes as smart as your collective service organization. And everyone has access to this intelligence, from your newest engineer to your most experienced support person.
Other differences between RevTwo AI and chatbots:
- RevTwo AI is well suited for service problems associated with complex machines;
- RevTwo AI does not require programming;
- When offering a solution, RevTwo AI considers the cost of a fix and how long it takes to perform;
- In the beginning, RevTwo AI can be field trained, learning as it goes along (much like an apprentice);
- RevTwo does not rely on NLP.
Are you looking at AI to help improve your service organization as well as your customers’ experience? Check us out at www.revtwo.com.