Photo by Markus Winkler on Unsplash

You might have read about the growing role of AI in the customer experience and customer support. Customer care is going through a revolution of new technology and we’re bombarded with amazing statistics such as this one:

“By 2025, as many as 95 percent of all customer interactions will be through channels supported by artificial intelligence (AI) technology”

But what do these AI buzzwords actually mean and how will they impact customer service? Here’s our guide to artificial intelligence terms in CX.

Artificial intelligence in customer support glossary:

‍AI

An AI is a machine that can perform functions that we associate with human intelligence. Until around 10 years ago, AI was most often associated with symbolic algorithms (logic) that did not use learning. Route planning algorithms and the first chess computers are "traditional" AIs. In recent years the lines have been blurred and the best performing algorithms all use some element of learning from data.

See also: Good, Old-fashioned AI & Real AI

Artificial Intelligence as a service (AIaaS)

AI as a Service (AIaaS) is the outsourcing of Artificial Intelligence by a third party vendor. Just as Software as a Service (SaaS) allows other organizations to take advantage of “off the shelf” systems other companies have produced without a huge investment, AI as a Service helps companies that do not have the capability or desire to build, test, and implement artificial intelligence systems from scratch. Companies like raffle are producing “plug and play” AI-driven solutions, such as AutoPilot and CoPilot, that are AIaaS.

Automation

Customer service automation in customer service is the process of implementing customer service tasks automatically, from routing issues to the correct teams to sending follow-up emails to customers. This eliminates manual labor using a variety of tools, including chatbots. Automation is not necessarily AI however, as simple tasks can be programmed without the need for machine learning. The majority of chatbots are automated, not real artificial intelligence.

Chatbot

A chatbot is built around a “decision tree,” which is designed to steer the conversation to an endpoint. The decision tree represents the logical flow of the interaction between the user and the bot. They can be effective tools for relatively simple interactions requiring binary choices, e.g. meat or vegetarian pizza. A chatbot may feature some natural language processing (NLP) capabilities but can’t learn from its previous interactions, so is not “Real AI.”

Read more: Conversational AI: to chatbot or not to chatbot?

Customer self-service

Customer self-service is a tool or service that allows users to find solutions themselves, often using a knowledge base. Organizations can use AI to direct customers to the right answer in that knowledge base, based on the question they’ve asked. For example, our AutoPilot product uses natural language processing to instantly find a relevant answer from companies’ knowledge bases and deliver them to the customer on their website.

Read more: 14 powerful stats that prove the importance of self-service in the customer experience

Data (Unlabeled)

Unlabeled data is one of the data types used to train language models. This can be text data collected from the internet or text available in companies. Practically unlimited unlabeled data is available, but care must be used, because a model could learn unhelpful or even offensive things.

See also: Language model

Data (labeled)

Labeled data is a type of data used to train language models. This could be a series of questions from historical customer query logs and labeled to help start the language model at a good level of performance before it begins to relearn (unless it’s a chatbot, which depends entirely on human input). 

See also: Language model

Deep Learning

Deep Learning is a subfield of machine learning that uses model architectures with many layers of computation inspired by the organization of the brain. The computation progresses layer by layer starting with input data and the final layer outputs the prediction. It is used for image classification, image generation (deep fakes), machine translation, speech recognition, speech synthesis, learning to play games and many other applications. 

‍Good, Old-fashioned AI? (GOFAI)

Good, Old-fashioned AI is technology that is built with “if-then” statements — rules programmed by humans. Sometimes known as rules engines or expert systems, they are useful for conducting repeated tasks but have little to do with actual intelligence. They automate processes but, unlike with “Real AI,” don’t self-learn or improve without human intervention. Most chatbots are built using this simple technology.

See also: Real AI


Knowledge base

A customer knowledge base is a collection of published documents and information, typically including answers to frequently asked questions, how-to guides and videos, troubleshooting instructions, introductory articles, glossaries and definition lists, and any other useful information for customers. Knowledge bases are made to help customers self-serve, without needing to ask for help from a customer support team member.

See also: Customer self-service

Language model

A language model can “understand” the meaning of sentences. Or, more precisely, if we take two sentences that carry the same meaning, then their representations will be similar. This currently requires training with labeled and unlabeled data. This is a very good foundation for customer service applications such as a question-answering system because it allows us to represent questions in a way that accurately reflects how customers ask them.

Read more: The science behind the raffle-lution: How language modelling has changed NLP — and our customer service products 

Machine Learning

Machine learning provides systems the ability to automatically learn and improve from experience without being explicitly programmed. This has the potential to improve customer service, customer loyalty and brand reputation, as well as enable employees to focus on higher value or more complex tasks. For example, a machine learning model can be trained on past customer tickets to assist the customer service agent find answers to new customer tickets.  

Narrow AI

Narrow AI is an intelligent system designed to perform singular tasks it is specifically programmed to do. This might be speech recognition, facial recognition, or driving a car. In the customer experience, Narrow AI is needed for effective social chatbots or human-robot interaction.

Natural Language Processing (NLP)

Natural Language Processing (NLP) is the use of computer algorithms for natural text. Machine translation, question-answering and dialogue systems are applications of NLP. A customer service query may be vague but, with NLP, we are able to interpret and analyze it. With deep learning NLP we can understand the words, sentences, and context of your customers’ queries. 

See also: Language model

Real AI

By contrast to Good, Old-fashioned AI, Real AI is more sophisticated technology which is able to self-learn. Using machine learning and neural networks, they require little to no human intervention. These programs alter themselves, are dynamic, and adjust based on the data they are exposed to. Thanks to this, they can assist customers based on previous interactions and user feedback. 

Read more: What is a real AI company?

AI: beyond buzzwords

When it comes to the customer experience, artificial intelligence is an exciting new development. However, it can be confusing. We hope that this glossary clears up some of your questions and explains the facts behind the buzzwords.

To find out more about how we’re using AI to revolutionize the customer service industry, read about our products: AutoPilot, CoPilot, and MissionControl.