If you haven't heard of chatbots yet -- or your experience is limited to novelty programs like Cleverbot -- chances are you'll be seeing more of them in the coming years. Why? Because companies are slowly starting to leverage chatbots as a way to manage basic communication tasks that used to belong solidly to the realm of human capabilities.
In this piece, Hristo Borisov, the Director of Product Management at Progress
, helps illuminate what chatbots are, how to build them, and their role in the future of business.
What Are Chatbots?
In short, chatbots are robots programmed to respond like humans. According to Borisov's definition, "A chatbot is a computer program that is capable of having a human-like conversation with a user by receiving and sending text messages for the purpose of automating a business process."
Chatbots can range from simple to advanced -- they fall on a spectrum of artificial intelligence. On the one end, says Borisov, you have "simple rule-based chatbots that can handle specific messages from users." These are less useful in a consumer setting because they require a specific input and are programmed to generate a specific output, so there's not much room for different conversational styles.
However, the more advanced chatbots (and the one's businesses will benefit most from) are grounded in AI, are capable of understanding conversational phrasing, and are programmed to actively learn from previous conversations so they're constantly able to grow and develop.
How Can Businesses Benefit From Using Chatbots?
Certain communication tasks will probably always require a human touch: marketing, PR, sales, more advanced customer service, and so forth. Where chatbots prove most useful, at least for now, is in the basic, less-skill-based areas of interaction.
At Borisov's company Progress, he uses the example of chatbots they built for hospitals, which "automate the process for patients to book doctor appointments by talking to a chatbot." These chatbots are powered by Artificial Intelligence and Natural Language Understanding, he says, meaning that "It is trained to understand different intents or conversations. Booking an appointment or contacting an operator are two examples of conversations that the chatbot can understand."
In practical terms, this allows businesses to increase their efficiency and reduce overhead. "The reduction of staff for repetitive processes requiring customer support employees is the biggest promise of chatbots in the long-term," says Borisov. "In one of our hospital examples, we are reducing the workload for the hospital's contact center by 30%, and patients can book appointments 24/7 without waiting for operators during peak hours."
What Is Involved In Building Modern Chatbots?
Here's where it gets technical.
Many basic chatbots are built using "functional programming," Borisov explains: "Developers today can use Natural Language Processing (NLP) algorithms to extract structured data from natural language, and use this information to create more intelligent chatbots. In combination with the power of NLP, developers use functional languages such as .NET or Java to create decisions-trees or slot-based algorithms that lead the user through a predefined conversation path."
However, functional programming has limits when it comes to automating conversations. Let's return to the example of Progress's hospital chatbots. "With functional programming], when you want an appointment, you will be asked first to provide a name of a doctor and then a date. However, this approach leads to two problems. First, a developer needs to implement in code all possible scenarios and inputs from the user, which leads to a huge codebase that is error-prone and expensive to support. Second, this approach feels like an automated machine to the user, since they cannot change their mind in the middle of the conversation, which is not natural."
The solution lies in using something called "declarative programming." Borisov explains: "In declarative programming, you describe what information you want to be extracted from a conversation instead of describing how by creating complex decision-trees. This way, a developer relies on cognitive flow algorithms that feel natural to the user and can handle scenarios that occur in natural conversation such as changing your intent at any point of the conversation. We have discovered that it's 16 times faster and twice as cheap to follow this approach."
How Will Chatbots Continue to Evolve?
There's still a lot of work to be done before chatbots will become a viable mainstream alternative to human-centric customer support methods. As the trend grows over time, "the developer tooling for creating engaging and intelligent chatbots will mature," says Borisov.
For the time being, he expects certain industries to have an easier time adopting them than others. "We see the adoption of chatbots to be led by healthcare and financial institutions due to their structured processes that are suitable for automation with chatbots and their increased investments in digital," he says.
Whatever happens with chatbots and AI in coming years and decades, it seems inarguable that the work landscape is about to undergo some massive changes. And that means it's time to get serious about making yourself an employee that machines can't replace.