What is Conversational AI?
Updated: Sep 14, 2020
If you've seen the phrase 'Conversational AI', you might wonder what that means. Is it just in reference to voice assistants like Amazon Alexa? What about Siri? Chatbots? Is it a fancy name for using a human voice for command and control instead of a graphical UI? And what's the Intelligence part? Is this AI supposed to be smarter than people? Does my business need it? How do I get started?
Over the past 6 years, devices that we talk to, like Google Home and Amazon's Echo, have experienced broad adoption. They enable convenience and even improved lives, while also causing frustration and privacy concerns. Like web and mobile technologies, these interactive voice systems are becoming a permanent part of our everyday lives.
Similarly, the rise in automated chat agent systems—chatbots—has been rapid and widespread. These also use natural language input in the form of text and visuals like buttons and emojis. Many of them are a terrible user experience, however they are growing in usage and presence in a number of beneficial ways.
To answer our questions, let's look at a breakdown of what's out there and what it can mean for your business and customers alike.
Where is conversational AI used?
As of 2020, Conversational AI means any system that meets two criteria: 1) humans interact with it primarily through spoken or written language, and 2) the system makes use of one or more AI / Machine Learning (ML) processes to identify spoken or written language and translate it to action. By that definition, Conversational AI systems are fairly common. Siri, Alexa, Google Assistant, Replika, Bank of America's Erica, and countless customer service chat and voice response systems all meet the two criteria. You've likely interacted with many of them, even if you weren't aware of it. You and AI won't have extended talks about complex human topics anytime soon, but the functional communication they support qualifies as a conversation.
Most systems featuring Conversational AI also use other interface modalities as well as ML for additional tasks beyond language recognition. Siri is embedded in iOS. The Echo Show displays images and information on screen nearly constantly and much of the functionality behind it also runs on ML. Other systems send email, present pictures, and offer buttons and other touch UI. These are used to start, facilitate, assist, and conclude conversational interactions. With this in mind, we see that Conversational AI is growing both as a standalone mode—e.g., voice in, voice out—as well as part of multimode offerings in which voice, text, touch, click, and image combine to enable a rich interaction.
So what's the Artificial Intelligence part anyway? Are these things supposed to be smarter than people?
The original goal of Artificial Intelligence was to reproduce computation that worked like our brains, meaning it could perform a variety of complex tasks using sophisticated reasoning. The research has been productive, but has not gotten very close to the goal. Now that goal is called 'General AI.' It's mostly a research hobby for computer scientists.
There's a subset of General AI that emerged though: a collection of technologies that have proven useful for business, science, and other endeavors. It's referred to as 'Narrow AI' or 'Specific AI.' Specific AI is made up of tools and methods such as machine learning (ML) algorithms and the abilities to harvest massive amounts of data for analysis and prediction in a constrained domain. Areas where Specific AI is proving to be a good addition to older technology include: customer service, healthcare, assistive technologies, and business analysis. Chatbots and virtual assistants exist because of Specific AI.
Much of the Specific AI we encounter is still in early stages of its potential. AI-powered customer service apps use voice and text recognition to handle a broad range of input, pull helpful information from thousands of documents, yet handle about the same range of core functions that their less advanced predecessors handled. The interactions are limited in context, primarily transactional, and often feel hard or confusing because they aren't built well.
More complex and nuanced interactions are not that useful and rarely result in good experiences yet. Voice assistants can tell us the weather anywhere in the world and play any song requested (and those aren't easy functions), but the usefulness is relegated to handling simple tasks one at a time a few times a day. Specific AI has also conquered human experts in the challenging games of Chess and Go. It's a fascinating accomplishment, but few of us are faced with needing the benefits of such specialized tasks to get work and life done.
Specific AI will become more useful; however, it requires robust work for interactions with it to feel effective and pleasant. Specific AI is very much a 'you get out what you put into it' technology. We help you identify the right ingredients and plan to execute on the potential.
Does my business need this? How would I get started?
Conversational AI interfaces produce business value for many companies large and small resulting in cost savings, sales, and more. There's no question that the business cases are there and we can show you why. As part of revenue enhancement strategies, these systems can enable new or expand on existing channels and lead to unexpected insights. Implementing them is more of a challenge than earlier interactive systems, though. To get the most out of them, you'll want to be informed and prepared. Our post 'Conversational AI: Three Keys to Getting the Most Out of Your Investment' tells you how to ensure a great start.
Looking to get going with your team on voice, chat, or multimodal apps? CCAI offers a 3.5 hour workshop for teams looking to accelerate the use of conversational AI.