Chatbots are computer programs able to carry out conversations with users without relying on a human agent. When chatbots are set up well, it should feel like you're talking to a real person.
You'll often see them on company websites, or on eCommerce stores — look out for messaging applications used to share information, manage customer queries and provide personalized customer experiences.
There are many types of chatbots out there today. How they are used depends on a number of factors, including:
Chatbots exist on many platforms. Some common platforms that chatbots operate on are:
Chatbots are used as a way to communicate with these platforms, navigate them, and even perform simple operations. One of Instabot's company philosophies is that chatbots are best leveraged inside existing applications, such as a website, mobile, email, or social media.
This allows a chatbot to access information already created, for example within a CRM system — chatbots can use powerful integrations to link up various parts of a business' marketing and sales operations, providing information for marketers as well as improving customer satisfaction and user experiences.
This means that chatbots are increasingly an important lead generation tool, able to both facilitate and record customer interactions for the purposes of marketing and data analysis.
There are many ways for a chatbot to operate, but currently three common ways are: rule-based logic, machine learning, and artificial intelligence.
Rule-based: When a bot is rule-based, it is given a simple decision-tree logic and algorithms for interpreting data. At its most basic level, it can be like a phone tree (if you want sales, press “1”), and in more advanced forms, for example in a healthcare setting, it can be used to diagnose illnesses quickly based on a variety of symptom data provided to the bot. Oftentimes chatbots are said to be “artificial intelligence” when actually their functionality is governed by rule-based logic. While some believe that decision trees are a subcategory of machine logic and artificial intelligence, we believe these are two very different things, and not to be confused.
Machine learning: A machine-learning bot is given information, and retains knowledge, which better prepares the bot for performing the same tasks or more advanced tasks in the future. In other words, it gets “smarter” the more you play and interact with it. Machine learning may be accomplished by technology called “neural networks,” which mimic processing information in the same way human brains perform tasks. A good example of very basic neural networks and machine learning is Google Quick Draw: the more people that doodle, the better the bot becomes at identifying the subject people are drawing.
Artificial intelligence: A bot truly operating on artificial intelligence is able to perceive the world, its multitude of variables, learn from experiences, achieve goals based on information perceived, and be economically viable — independent of human interaction. Artificial intelligence is a vast science which can break down into a multitude of sub-categories, such as reasoning, learning, natural language processing, among many others.
Natural language processing (NLP) or natural language understanding (NLU) is often a first step in employing artificial intelligence in chatbots. This is essentially allowing a computer to both speak and understand humans as they actually speak, in terms of dialects, colloquialisms, word choice, etc. Most chatbots use NLP in the form of “intent-based” systems, which is just allowing a chatbot to understand a user's intent and to therefore deliver an appropriate response. For example, in order to ask someone how they are doing, you could use the following phrases, “How are you doing?”, “Wussup?”, “How goes it?” “'Sup, dude?”, or “What's cooking?”. Effective NLP/NLU will allow a chatbot to understand that all of these phrases essentially mean the same thing, or have the same intent.
The logic of the bot you choose may depend on your goals. Further, you can use a combination of all these methods with your bot. While there have been many advances in these systems in the last several years, machine learning and artificial intelligence are still in their infancy. Hence, the most advanced capabilities are not available to the public, and/or may be inconsistent in terms of accuracy, based on inputs and how they are used.
Chatbot methods of communication
Currently, chatbots communicate through text (as in Facebook bots), meaning you type in your inputs/responses, or through audio (talking to Amazon Echo).
The type of communication you choose for your bot will depend on circumstances of use. For example, voice-controlled bots work well in quiet places such as the home or in your car, where the bot can easily recognize your words and where it does not impact other people around you.
However, if your bot is for the office or use in public places, then your bot would best be served through text, where multiple interactions can be handled one-on-one, without conflicting with someone else's space or with their own bot experience.
With so many variables, there are infinite types of bots. You can have a Slack machine-learning bot which operates through voice, or a Facebook rule-based bot which operates through text. How you build your bot and its interface depend on what your users will need from the experience and how it will be used day-to-day.
Chatbots are a fantastic resource for scaling customer support, as they reduce the need for service agents while giving customers valuable answers in real time. Having the technology to solve customer problems and reduce wait times is a very important tool for many businesses.
Businesses use chatbot technology to interact with potential and existing customers, giving them a personalized experience when using a tool or a website. Personalized experiences are very important for driving sales and also help with customer retention.
eCommerce businesses rely on customer experience — buyer journeys need to be made as smooth as possible. Chatbots in this context can be used as a kind of virtual assistant, helping to streamline the customer journey towards conversion.
Chatbots make an experience or task easier. For example, if you want to call a taxi, you can go down to the street and stand on the sidewalk with your hand in the air, waiting for one to drive by and flag it down, or pick up the phone to request it one hour in advance.
Or you can go to your Amazon Echo “Alexa” chatbot, and ask “her” to call you an Uber, and you'll have a car service show up to your house in minutes.
You can have your parking tickets quickly overturned with the Do-Not-Pay bot, or easily order a pizza from the Domino's Pizza bot. Chatbots can make using websites and apps easier, simplify online searches, request services, notify you of information that you requested when it becomes available, and so much more.
The ability to automate self-service has become extremely valuable, especially in the context of trying to find information or make a purchase online.
Chatbots optimize virtual experiences, for the benefit of the business and consumer, and as the technology behind them improves they'll become even more integrated into our daily lives.
Read our blog How Do Chatbots Work? to find answers to more FAQs about chatbots and chatbot technology.
Here is a mini-history of chatbots (and all things AI) from 1950 to today, and then some.
1950: Turing Test
Alan Turing publishes Computing Machinery and Intelligence, introducing the concept of the Turing Test which tests a machine’s ability to exhibit behavior indistinguishable from that of a human.
1956: Dartmouth Conferences
The Dartmouth Summer Research Project on Artificial Intelligence is held, establishing AI as a field of study. Organized by Assistant Professor of Mathematics John McCarthy, the conferences lasted approximately 6 to 8 weeks.
ELIZA, an early natural language processing computer program that simulated a Rogerian psychotherapist, is created at the MIT Artificial Intelligence Laboratory by Joseph Weizenbaum. It is said to be one of the first programs able to pass the Turing Test.
PARRY, a chatbot created to simulate a person with paranoid schizophrenia, was created in 1972 by psychiatrist Kenneth Colby. In the same year, PARRY and ELIZA “met” and “talked” to each other at the International Conference on Computer Communications in Washington D.C.
1980: Chinese Room
The Chinese Room argument is introduced in philosopher John Searle’s paper “Minds, Brains and Programs” published in Behavioral and Brain Sciencesscientific journal. The argument poses a hypothetical scenario which implies that a computer has no consciousness, no matter how human-like it may behave.
Racter (short for Raconteur), an AI computer program written by William Chamberlain and Thomas Etter, is revealed in a book called The Policeman’s Beard is Half Constructed. Racter is said to have to have authored the book, entirely.
1992: Dr. Sbaitso
Dr. Sbaitso, an AI speech synthesis program for MS DOS-based PCs is released. Although the program attempted to resemble a real-life psychologist, the use of a digitized voice and its repetitive responses made it feel otherwise.
Artificial Linguistic Internet Computer Entity, or A.L.I.C.E for short, comes to life. Heavily inspired by ELIZA, the natural language processing bot had the ability to engage in a conversation with a human by applying heuristic pattern matching rules to the human’s input.
Jabberwacky is created by British programmer Rollo Carpenter as one of the earliest forms of human conversation based AI. Built mainly as a form of entertainment, Carpenter also intended it to be capable of passing the Turing Test.
SmarterChild, the intelligent bot developed by ActiveBuddy, Inc. is released across popular instant messaging and SMS platforms. The robot sat on the AIM buddy list of millions of kids and adults across the globe until the technology was shelved following an acquisition of the company by Microsoft.
GooglyMinotaur, an AOL Instant Messenger bot, was developed by ActiveBuddy to promote Radiohead’s fifth album, Amnesiac. Its release marks one of the first instances of bots used for commercial means. After conversing with nearly one million people about Radiohead related content, the bot was switched off. The cause of death is undetermined.
2006: IBM’s Watson
IBM’s supercomputer Watson, named after the company’s first CEO, is developed with the ability to answer questions posed in natural language. In 2011, the computer competed on the game show Jeopardy beating former winners Brad Rutter and Ken Jennings. To this day, Watson powers countless businesses across different industries.
Siri, Apple’s intelligent personal voice assistant, is developed by Siri Inc. and released as a standalone application. After being acquired by Apple that year, the program was integrated into iOS, with the ability to interact with a number of Apple’s default applications. Today, Siri can be used across apps within Apple’s iOS, watchOS, tvOS and macOS.
2012: Google Now
Google now, Google’s own intelligent personal assistant, is released for Android. Using a natural language user interface, the bot can answer questions, make recommendations and perform actions across various web services. In 2016, an evolved version of Google Now which hosts the ability to engage in a two-way dialogue, called Google Assistant, was announced.
2015: Amazon’s Alexa
Amazon releases their own intelligent personal assistant, Alexa, which is capable of performing countless tasks through voice interaction. Alexa is the operating system found within the Amazon Echo smart speaker, which acts as a home automation hub by controlling numerous smart devices.
2015: Microsoft’s Cortana
As a nod to the fictional AI hologram character in the Halo video game series, Microsoft releases their own intelligent personal assistant, Cortana. Available in numerous languages, Cortana serves as a key ingredient of Microsoft’s operating systems “makeover.”
Microsoft releases an intelligent chatbot, named Tay, on Twitter under the handle @TayandYou. Designed to mimic the language patterns of a nineteen year old female and learn from interacting with Twitter users, Tay soon became known as “The AI with zero chill” as she began to exhibit offensive behavior. It was taken down only 16 hours after its launch.
2016: Betaworks Botcamp
Betaworks announces a 90-day pre-seed program for chatbot startups. Ten companies are accepted, each receiving $200k and an office to work out of at the Betaworks Studio space in NYC.
2016: Facebook / Slack / Telegram / Kik / Apple Bots Launch
Bots are everywhere. Facebook announces a platform for building bots for Messenger, and tens of thousands of them are created within months. Other messaging services, like Slack, Telegram and Kik do the same. Later on in the year, Apple opens up iMessage to third-party developers. Bots have officially arrived.
2016: ROKO Labs Instabot
ROKO Labs launches Instabot, a platform for creating your own chatbot that you can launch on existing mobile apps, websites, and in emails.
Want to know more about other current bots? You can check out this handy directory, Botlist.