Tech Zone

Hey Alexa! How do you Work?

Chatbots are firmly embedded in customer service systems today.

Despite their growing use, they remain a bit of a mystery as to what they are, how they work, and their future.

Log into an account today that has a messaging function, ask a question or seek assistance, and in many instances a chatbot will introduce itself and ask how it can help. Chatbots are arguably the first mainstream use of artificial intelligence (AI) to provide general customer services. Sure, there are vehicles with AI technology, but there are more vehicles without it. Increasingly, chatbots are used by big and small companies to speed up customer service, deliver improved customer service, and reduce labour costs.

Despite their growing use, chatbots still have a bit of mystery about them for non-technical people. As information technology (IT) experts strive to develop chatbots that are as human as possible in terms of communication and problem–solving abilities, everyone else wonders what they really are, how they work, and how they help businesses now and are expected to help in the future.

Mixing it up
Various chatbots are a mix of technologies that include AI, machine learning (ML) and natural language processing (NLP). Many brands are already using them with the goal of improving customer services, increasing sales and streamlining consumer problem resolutions.

The Sephora chatbot on the messaging app Kik is one of the most touted examples of successfully using a chatbot to interact with current and potential customers. The beauty chatbot, aimed at Gen Z, is a shopping assistant that offers tips and tutorials for users. It helps users find products, suggests the best products for their purposes, and shares reviews on specific products.

The Sephora chatbot is an assistant chatbot. The AT&T chatbot is a support chatbot that knows a tremendous amount about the AT&T company and can answer questions or direct customers to the appropriate person. There are also skills chatbots like Alexa that can follow commands. The bottom line though is that chatbots are problem solvers, whether looking for the ideal mascara, asking the lights to be turned on, or getting help with a TV receiver that is not working right.

A chatbot is a software program with an architecture that combines a set of technologies, especially ML and NLP, in order to have conversations with humans and produce results that mimic human thinking. As the chatbot technology advanced, it also embraced AI which enabled the processing of more complex activities.

A bot (as in robot) is the computer's ability to understand human speech or text, and a chat robot (chatbot) simulates human conversation. The chatbot can respond to voice commands or text.

Learning From Experience
The chatbot responds to human inquiries by drawing on a knowledge database that grows as the computer learns by itself through its experiences. If the chatbot is unable to answer a query or respond to a statement due to lack of programming and learning, it will usually pass the person on to a human.

However, it learns from that interaction, continuing to grow in learning and scope. The sophistication of the chatbot depends on the programming and the data it can access. The highest quality customer services chatbots, for example, have access to billions of data bits from past customer calls.

The chatbot architecture is complex in every situation due to programming requirements. The basic chatbot is fueled by rules programmed into the software. Not all chatbots use AI, though that is likely to change.

Today, many chatbots interact with humans via a chat interface. Sephora uses Kik, but other familiar chat interfaces include Facebook Messenger, SMS, and WhatsApp. Chatbots can also communicate via smart home devices and smart speakers.

Chatbots cannot solve all customer issues. They are ideal for answering common questions or requests like “Reset my password” or “Do you have this product available?” These are clear-cut simple communications that chatbots can easily answer, freeing up the customer service representative's time.

When a human communicates with a simple rule-driven chatbot, the program parses the input into keywords and then scans the database for references to the same keywords. There is a set of predefined rules. For example, a customer expresses interest in shoes, so the chatbot asks questions to help narrow the product search and make recommendations. However, a chatbot in its simplest form does not recognize natural language. Since it only recognizes keywords, but not context, the responses may not always make sense.

Chatbots That Understand and Interpret
Becoming more common are chatbots driven by NLP or natural language understanding (NLU).

NLP is a branch of AI, and it is exactly what it sounds like – computers understanding, interpreting or analyzing, manipulating, and generating language the way humans do.

The ultimate goal of programmers, of course, is to develop software that can talk and respond just like humans.
NLU is related to NLP and thus also a branch of AI. It is computer software that understands human input in the form of sentences in text or speech.

NLU enables chatbots to better simulate human conversation within context, manage a dialogue and make appropriate responses. When the chatbot is powered by AI, it can learn from each interaction with a human.

AI, NLP and NLU make the chatbot more sophisticated in terms of its ability to give appropriate responses. The ultimate goal of programmers, of course, is to develop software that can talk and respond just like humans.

Still Need Humans
There are many business uses for chatbots. They are used as call centre initial contacts and digital personal assistants. They are also increasingly used for back-end systems, like inventory management and providing assistance with travel plans. Businesses can add voice or text chat interface to create chatbots that help customers perform tasks like accessing a bank account or scheduling Uber.

The current limitation is that chatbots are not forward-looking. In other words, they rely on past information and cannot anticipate the human's plans or intents. They are very specific.

There is likely to be many more uses as the technology advances, and there is no doubt that more companies will use chatbots. On the other hand, chatbots will never totally replace humans.

Already, frustrated consumers are directed to human representatives by chatbots, proving once again that technology always has limitations and humans will always be needed when talking about customer service.