Demystifying the Virtual Agent vs Chatbot Debate
Posted by on March 07, 2019
Matt Klassen is the vice president of product marketing at Cherwell. He is passionate about enabling enterprises to accelerate their digital journey through better software and better service. Matt has 25 years experience in developing, architecting, selling, and marketing enterprise software solutions for IT and product teams.
Artificial intelligence (AI) is playing an increasingly prominent role in technological change across industries, especially in industries that can leverage the technology to deliver customer service functions. Although chatbots have been around for decades now, the idea of a virtual agent that can seamlessly interact with customers, provide relevant information, and even complete basic functions like booking an appointment or scheduling a call-back is relatively new.
On top of that, there seems to be a general confusion surrounding the difference between artificial intelligence robots, chatbots, and virtual agents—aren't they all really the same thing?
The truth is that with the rapid development of AI technology and an increasing number of companies marketing proprietary AI products, it's getting more difficult to keep up with the appropriate terminology and evolving capabilities of web-deployed bots. In addition, most tech vendors want to market their product as "the next big thing." That means it's important for you to understand what technology you're buying into and what functionalities to expect—regardless of whether it's branded as a "chatbot" or a "virtual agent."
To help you make the best AI investment for your business, we've spent the last week demystifying the virtual agent vs chatbot debate and trying to answer three questions:
- Are virtual agents and chatbots the same thing?
- If not, how are they different?
- What technologies are driving virtual agents and chatbots in 2018 and when should each one be deployed?
Here's everything we learned so far.
Bots, Chatbots, and Virtual Agents: What’s the Difference?
As with any discussion about definitions, it's important to start with the caveat that not everyone may necessarily agree on how these technologies are distinguished from each other. Language should be descriptive, not prescriptive—the point isn't to tell other people what to say, but to create more commonality around terminology so that we can all communicate with each other more effectively.
To understand what a chatbot is, we should begin with a short discussion about bots. The term "bot" is short for robot, but we're not talking about robots like the maid from The Jetsons who can dust your wardrobe. The bots we're talking about are automated programs that can be deployed via the internet, and there are many different kinds. Search engines use a special kind of bot called a web crawler, or spider bot, to find and index web pages so they can appear on search results. There are even malicious bots called spambots that scrape the internet for email addresses for inclusion in spam campaigns.
A chatbot's function is entirely suggested by its name—it's been programmed for chatting, conversation, or communication. If you've never experienced conversation with a chatbot, there are several free online chatbots that can entertain you for hours. There are also several websites that will allow users to create their own scripted chatbots that can be deployed on the web or on social media.
Automated bots are typically run using scripts—little bits of code that provide basic instructions for what the bot should do. Some bots, including chatbots, are not run automatically and only execute a command when they receive a specific input. Chatbots that deliver scripted responses to pre-determined inputs are an example of this type of bot.
When we talk about virtual agents, we're talking about something slightly more advanced than a chatbot. As its name suggests, a virtual agent should be able to do more than deliver a scripted response during a conversation. Just like with a live agent, it's more likely that a virtual agent would be able to accomplish additional tasks like resolving common issues such as password resets, performing automated tasks related to know requests, or escalating the interaction to a live help desk agent.
Chatbots and virtual agents are similar in that they interact with the customer to deliver customer service. What really makes the difference is the technologies underlying these applications and how that translates into capabilities that can benefit the customer. In the next section, we'll talk a little more about those technologies and what kind of service they can help deliver.
Chatbot & Virtual Agent Technology Explained
While it's true that chatbots are most commonly interacted with through an instant messaging tool rather than as part of a live conversation, the key differentiator between chatbots and virtual agents is the technology that underlies them. Although the first chatbots were invented decades ago, changes in technology over the last 50 years have had a profound effect on what users experience when interacting with a chatbot.
The First Chatbots Used Scripted Responses
The first chatbots ever created were programmed in the late 1960s and early 1970s. These simple bots used scripted responses to communicate with users based on the user input, and could be programmed to respond to certain key words or phrases. Still, the need for scripted responses imposed severe limitations on what these chatbots were capable of. They could only give responses that they were programmed to give, so there was no programmed learning process and no sense of the bot saying anything original. Early chatbots were also characterized by nonexistent contextual understand or memory—while the bot might respond appropriately some of the time, it would immediately "forget" relevant information in the conversation.
Chatbots Today Use Natural Language Processing
The chatbots that are programmed today make use of a much more sophisticated virtual conversation technology known as natural language processing (NLP). This technology simulates AI by allowing a chatbot to more easily understand and interpret instructions or input in natural language, allowing the user to speak to the chatbot just like they would a human being. In addition to the traditional methods of scripting responses, NLP gives the chatbots of today additional flexibility and capacity when it comes to interpreting commands and responding appropriately.
AI Chatbots Benefit from Machine Learning Across Interactions
Today's chatbots also benefit from learning capabilities based on machine algorithms, also known as machine learning (or ML). Machine learning is a cutting-edge technology that is in many ways still in its infancy. Still, this is the technology that allowed Google to train its Alpha Zero Chess Engine into the strongest chess computer in history in a space of just six hours—by feeding it with all of the data on chess that ever existed.
An AI chatbot or virtual agent can train itself based on transcripts or recordings of previous customer service interactions. Unlike older chatbots that were programmed to deliver a canned response for each input, these bots can learn to manage individual customer interactions cohesively from start to finish. They can greet the customer, offer service, provide information, satisfy a request, confirm the customer's satisfaction, and end the interaction, just like they "learned" from studying other customer interactions.
What has changed significantly is the technology and software that underlies the bots themselves, with the newer bots demonstrating significantly advanced capabilities and offering more useful conversations and interactions.
What’s in a Name? Separating the Functions of Chatbots and Virtual Agents
After all that, are we any closer to demystifying the virtual agent vs chatbot debate? What we ultimately discovered was that chatbots have been around for over 50 years, and they've always been called chatbots regardless of the software and functionalities that they were built around. Chatbots and virtual agents may be similar in their programming, but the way that they function is dependent on the context.
The primary function for a chatbot is to facilitate a useful conversation with the user. In a customer service context, that could mean receiving a question from the customer, processing the question using NLP, searching an accessible knowledge base for the appropriate information, and supplying that information as a response. If we're thinking about a chatbot for IT service management, it should be able to understand the customer query, search for relevant information in the IT organization's knowledge base, and supply the information.
Once we start talking about virtual agents, however, the expectations start to change in regard to what the bot should be capable of. If a bot is going to be called a "virtual agent," it can't simply provide answers from a knowledge base. After all, an agent is someone that can act on your behalf, so a virtual agent needs to be able to do something for you. Virtual agents would also be more likely to have some kind of machine learning component, helping them better understand and satisfy customer requests as they gain experience across multiple interactions.
The ability to understand and keep track of context in conversations is a relatively new feature for chatbots. Some intelligent virtual agents can remember information from the user, apply context to generate more realistic conversations and decide when to escalate to a human agent or secretly ask a human for help. When we refer to a virtual agent, we’re referring to a bot that is able to discern between an incident (something is wrong) and a request (something in a service catalog) and help the user not only create a ticket, but in some cases resolve the ticket or request without involving a human.
Think about how a virtual agent vs chatbot would respond to your customer query "I can't access my email."
A chatbot might notice the keywords "access" and "email," then search the knowledge base for information about accessing your email and provide the information. If that's not helpful, you might have to escalate the issue yourself by calling a call center to speak with a human agent.
A virtual agent with machine learning might know that 68% of the time when someone says "I can't access my email," what they want is to reset their password. The virtual agent would immediately offer the user the option of resetting their password using their back-up email address or security questions. The virtual agent could even complete the request and verify that the user can access their email before closing the interaction.
Virtual agents can be programmed to do many of the same things that a human IT agent would do during a customer service interaction. Asking questions and responding in a courteous way, verifying what the user wants before performing a service, and even consulting the organization's knowledge base or service catalog for additional information are all functions that can be performed by a virtual agent. Today's virtual agents can even consult a human agent behind the scenes when required, ensuring a seamless customer experience for users while reaping the benefits of automated service delivery.
Whether your IT organization uses a chatbot or virtual agent depends on the level of service you wish to offer, but either way, it's important to lay the proper groundwork that will jump-start your AI initiative.
The terms "Chatbot" and "Virtual Agent" are often used interchangeably, but it is important to establish a common terminology so we can communicate more effectively about these developing technologies. Both chatbots and virtual agents leverage varying degrees of software technology to deliver an experience or service to the user. While chatbots are used for conversations and communication, virtual agents are typically expected to act in the role of an "agent" and perform at least some services for the user. To facilitate this expectation, virtual agents often contain more sophisticated software, such as deep neural networks, machine learning, or natural language understanding, that allows them to better understand and interpret the needs and desires of the customer.
Interested in what Cherwell can do for your organization? Request a demonstration today!
You might also be interested in
Blog 10 min
ITIL Service Desk Responsibilities
Implementing an ITIL service desk can be intimidating when you first get started. Try these strategies.
Blog 2 min
Most IT Orgs Have Embraced AI—But There's a Catch
Dismiss any impression you have of artificial intelligence (AI) being the technology “of the future.” According to a recent study we conducted in partnership with global research firm IDG—among 260 IT decision makers at organizations over 1000 employees—the future is now: a whopping 71 percent of IT organizations have already implemented one or more AI projects.
5 Powerful Self-Service Features That Will Shift Your Service Delivery into High Gear
When your service portal meets the needs of end users, your IT team can devote itself to more challenging issues and initiatives.