Back to basics: What is a chatbot and how do they work?

Today, it seems as if chatbots have become an intrinsic part of our everyday life. We use them to complete a variety of different tasks – whether it be to make a booking, ask a simple question or even order a pizza! By imitating the fluidity of human discourse, and with their ability to rapidly learn, they have had a huge impact on our lives. While we have discussed a variety of chatbot related topics in the past, this article will take it back to basics. First off, how do you actually define a chatbot? And how exactly do they work?

What is a chatbot?

Chatbot is short for “chat robot”, and it is an artificial intelligence (AI) system designed to carry out a conversation with a user. They usually appear in the form of a messaging or “chat” interface for example Kik or Facebook Messenger. Chatbots provide answers to general, straightforward queries, that can be answered without the need for a human customer service representative.

Chatbots have been in existence from as early as 1966 and have since been heavily developed and modernised over the years. They can now be found in a variety of places, including websites, emails, messaging applications and social media channels. You may have been on a website before and seen a chat box pop up in the bottom right corner accompanied by the phrase “Can I Help You?”. This is one of the many forms in which chatbots can appear, some of which are purely a stand alone bot, but some are linked to a live chat service which will seamlessly route you to a human agent if necessary.

To further improve user experience, the discourse of a chatbot is designed to sound as much like a human as possible. This is the beauty of the chatbot – you tend to forget that you are talking to an automated system rather than a real life assistant. By adopting a natural and conversational tone chatbots can, in most cases, provide the user with the same assistance as that of a human.

How do chatbots work?

Now that you understand what a chatbot is – I bet you’re wondering how on earth do they work? A chatbot is developed using a combination of knowledge-base, ​Machine Learning ​and Natural Language Processing (NLP). These chatbot components allow for the translation of slang, dialects and incorrect spelling, whilst isolating keywords which are tagged as intents or topics of conversation. The chatbot is able to use these processes to connect the natural language that a person uses in everyday life, to the stored answers held within a knowledge base.

So how does this work exactly? Taking an example from our Council clients, a constituent may type the following question into the chatbot:

“I want to buy a parking badge”.

The chatbot would recognise ‘want’ and ‘buy’ as the intent and ‘parking badge’ as the topic. It is able to establish the users requirement and verify if the chatbot has a means of buying a parking badge. In this scenario if there is a single parking badge available to buy, the chatbot will return either a link to the correct location or link to an external system and allow the constituent to register for their new parking badge directly from the chatbot.

The complexity increases when a chatbot has two possible options available in the knowledge base, in this example a constituent can be interested in a blue parking badge (disability) or a council parking badge. With multiple options available how does the chatbot decide and handle the multiple responses?

There are two distinct methods for dealing with this scenario. The simplest method is to present both parking options to the customer in the form of a quick reply (button). The customer then clicks the option they were referring to and they are directed to the right area. The second option is via a weighted response process in which the company who are managing the chatbot decide the highest priority. If they think the customer should be presented with the blue badge response when the chatbot is conflicted, then this option can be presented to provide a seamless experience. A cleverbot can be fine tuned to combine both methods and following a series of quick replies take the most selected option to define the highest priority response.

Weighting can additionally be used to provide a “preferred” response, when a customer asks two questions in the same sentence. This doesn’t happen as often as you may think. We have become conditioned to talking to robots, as they have been integrated into our lives at an exponential rate over the past few years. Amazon Alexa and Google Home devices are a prime example of this, as they have exceed total sales of 100 million this year.

Despite our adaption to working with robots, chatbots need to handle all eventualities and weighting plays a major part in deciding on the “most important” questions to address. Let's take a look at a two question sentence:

“What time can I park here until and how can I buy a parking badge?”

The chatbot will decide through one of the weighting methods, described above, if the parking times or the purchasing of a parking badge takes priority and will adjust the response accordingly. Receiving a response to one of the two questions, often results in a happy customer and they will naturally proceed to ask the second question, now that the first has been resolved. This is not too dissimilar to how a human conversation would work, as it would be unlikely for you to be able to address both questions with a single response. The next step in the chatbot evolution is to not wait for the customer to ask the follow up question, but instead cache the secondary question. At the point of completion of the first issue, the chatbot would then be able to present the answer to the second question, without being prompted further.

Machine Learning:

In the early stages of a chatbot implementation, it is likely that the bot will encounter some questions that it hasn’t yet been programmed to answer. Chatbots can handle an unknown response in a number of ways, such as a default message or handing over to live chat for a human to takeover.

In traditional development new responses would require a developer to code the new response and plug the gap for the future. With the adoption of machine learning in chatbots, data can be used to derive the chatbots responses and essentially train the chatbot, making the responses more detailed and focused overtime. A chatbot can become increasingly knowledgeable and with enough data could can reach the point at which a chatbot does not require human intervention.

Integrations:

Chatbots are relatively new technology but they have come a long way in a short space of time and have evolved to integrate with customers existing infrastructure. Queries can be enhanced by providing a personalised response rather than that of a generic knowledge base.

Utilising the ever growing use of APIs (Application Programming Interfaces), information can be retrieved instantaneously and seamlessly. For example, if we look at integrating with a Council's available APIs, tailored responses can be retrieved in regards to a customer's refuse collection days, local Councillor details and nearby services. Customers can also pay their council tax within the bot using secure authenticated payment systems or even report a pothole in a specific area. The bot can then analyse these reports and aggregate multiple reports by location and topic, making it easy for Councils to identify key priorities without having to sift through hundreds of individual reports.

How can a business benefit from chatbots?

By using chatbots, businesses can massively reduce the need for a human customer service team. Nevertheless, most businesses will still need some form of human customer service for the more complicated queries that the chatbots are not yet able to answer. Having said this, chatbots are still incredibly useful in answering many of the simple, straightforward questions that don’t require any human intervention. After all, there is little point in hiring a human to answer the basic questions that a chatbot can answer instead.

As well as a limited need for a workforce, chatbots allow businesses to be able to offer customer support on a 24/7 basis. No matter what time of day it is, a customer is able to have their query answered – which is not the case with customer service helplines that are only open during specified hours of the day.

You also improve customer experience and reduce wait times by increasing first contact resolution. This especially applies to companies who receive a large volume of customer queries on a daily basis. While those calling help centers or using online live chats may have to wait to be assigned a customer service agent during busy periods, a chatbot can simultaneously maintain conversations with thousands of different people.

Lastly, the use of chatbots allows companies to incorporate customer service into popular messaging platforms. As of 2017 Facebook Messenger had 1.3 billion users, so it’s hardly surprising that many companies have started to incorporate chatbots into their social accounts.

Through doing this, businesses can provide their customers with easy access customer service on the platform where they are likely to already communicate with their friends and family.

Hopefully you know have better understanding of what a chatbot is and how they work. Chatbots are revolutionising the customer service industry, and pose many benefits to businesses in their ability to replicate human discourse and answer customer queries. If you are interested in having a chatbot built for your business, then b​ook an IntelAgent demo​ today.