A while ago lots of people were saying chatbots were going to revolutionise the online world, but most of them turned out to be irritating and not all that useful. It seemed like the fabled chatbot era was over before it began. Since those ‘dark’ times, a lot has changed and the technology has finally started to catch up with the hype, with chatbots (or conversational AI) now able to be used across webchat, messenger services and using voice through voice assistants and traditional phone call IVRs.

So What Are Chatbots?

Chatbots are essentially a service that users interact with via an interface, often appearing as a pop up on the side or bottom of a website, in a messenger or other mobile app. The good ones are focussed on helping people with questions about products or services or to complete tasks or transactions. All of this is done with the aim of taking the more mundane, repetitive processes off the customer service teams so they can be better utilised to deal with more complex issues that machines can’t manage.

Driven by Artificial Intelligence (AI), they seek to mimic human interactions without trying to pretend to be one. They provide real-time interaction with your business any time of the day or night , enabling businesses to service more people more efficiently and, if done right, improve customer satisfaction at the same time.
 

 
Using a chatbot to book flights. Users input their data like destination and dates, and the chatbot returns relevant results. Image source.
 

Brief History of Chatbots

Since the 1960s people have attempted to create an intelligent human/computer interaction based on a natural language model - a model based on interpreting and understanding language as humans use it, based on typing or speaking in plain English. The idea has permeated pop culture for a generation or more from Hal in ‘Space Odyssey’ through to more recent times like the character in ‘Her’ that Joaquin Phoenix falls in love with.

One of the original (and probably the most famous) bots was ELIZA - a chatbot therapist who offered some basic advice to her “patients”.
 

 
A ‘conversation with ELIZA’. From the 70’s to 2010 there were numerous other attempts to master the domain. Strangely, most of these bots always offered medical advice or assistance. Image source.
 
Then in 2010 Apple gave birth to Siri. Siri is not classified as a chatbot but rather a ‘personal assistant’. The technology is widely regarded as the catalyst that spurred the rapid growth of the Natural Language and AI bot landscape.


How Do They Work

A chatbot is regarded as the simplest form of conversational AI, which allows a computer to understand, process and respond to text and voice. A very basic model will use a template and ‘listen’ for specific keywords. In this simple model, if the ‘right’ keywords aren’t used, then the chatbot can’t effectively interact as it has a limited number of outputs, based on inputs from users.

More advanced chatbots use Natural Language Processing (NLP) or Natural Language Understanding to help computers comprehend requests and allow them to respond within the context. Chatbots using NLP understand the context and meaning behind a query, even when it doesn’t use a specified set of keywords. These chatbots also use Machine Learning (ML) to train their knowledge base to better understand topics and areas of expertise, allowing them to become more helpful to users as time goes on.

A virtual assistant like Siri will use a more complex conversational AI model. The assistant should be able to understand and respond to basic conversations and understand the nuances of the conversation. The first complex interaction is the language model and understanding the request, and the second part is to identify the action that needs to be taken to fulfill the request.
 

What Can Chatbots Do?

The basic use case for a chatbot is to make it easier for the user to interact with a business to accomplish a task. For example, you may interact with a chatbot to help you with a support request on a website to change your address details instead of waiting for a support agent on the phone. Chatbots can execute repetitive, mundane requests at scale to allow human staff to focus on more complex requests.

Depending on the implementation chatbots can also hand off conversations that they cannot assist with by seamlessly transferring the information you’ve already provided to a support agent that can continue the conversation. This still saves support staff time and if handled well does not cause any friction with the customer or prospect.

The power of a chatbot depends on two key criteria – the systems it is connected to and can interact with and the extent of the tasks it has been trained to handle. That said, one of the best features of a chatbot is that it can grow its capability over time, you can complete an initial roll out with a limited scope of tasks and integrations for a low price point then expand the use cases once the advantages of the platform become clearer to the business.


The Omni Channel Experience

In an age where clients expect immediate responses to their requests, chatbots can assist across a number of interaction points.

A number of chatbot platforms can offer a true omni channel experience utilising a single core chatbot that can interact via website chat, Facebook Messenger, Virtual Assistants (like Google or Alexa) and also over telephone using text to voice and voice to text technologies (using a service like Twilio Voice API).

This approach provides efficiency in development of the core conversation flows (create it once and use it across multiple channels) whilst also allowing for branching with different variations of language or response across each channel.
 

Challenges

Although the AI driven chatbots have improved exponentially over the past few years, they still require some work to keep up normal conversations. It is difficult for these models to interpret the tone of a conversation and fully understand the context of every sentence a human can throw at them. Some language models add in frustration detection protocols for the requests received and, based on a threshold, decide whether or not to handoff to a human to complete the conversation with the user.

It is also best practice to not attempt to hide the fact that people are interacting with a bot. There are a group of users that will prefer to interact with a human and refuse to chat with a computer – and it is far better if you don’t try to trick this cohort. You can always set up the chatbot to hand off to the correct human assistant for the group of people that prefer this interaction. If users aren’t forced into only using a chatbot then this is a fairly simple challenge to overcome.
 

Chatbots are evolving rapidly and becoming more powerful, with use cases extending far beyond the initial applications that had been imagined 50+ years ago. Organisations can leverage them as a business tool alongside traditional front-line staff to provide customer support and sales assistance. In an age when visitors are trying to connect with brands when it suits them, having an effective AI chatbot to guide and support these interactions means brands can connect with more users in helpful ways no matter the time of day.

You may also like

A Simple Guide To Getting Started with Your Kentico Xperience Audience Segmentation

Part one of a three part article series focused on empowering you to take the leap into Kentico Xperience (formerly Kentico EMS) and start creating online marketing campaigns. In part one, we focus on yielding a meaningful database of website vistors by utilising Kentico's segmentation tools.

Keep Reading

BERT, Content & SEO

BERT was slated as one of the biggest changes to Google's algorithms in years. What is BERT and how can organisations create content to perform well in Google in a post-BERT world?

Keep Reading

Newsletter sign up

Every couple of months we send out an update on what's been happening around our office and the web. Sign up and see what you think. And of course, we never spam.