Chatbot: What Is It and How to Build One
⏱ Reading Time: 16 minutes
Time’s person of the year 2006 was “You”. The digital presence anonymously contributing wikis, YouTube, MySpace, Facebook, and the multitudes of other websites with user-generated content. Emerging digital presence, chatbots who are yet to be explored and perfected, in that sense should be a Time’s person of the century. Bots have become an integral part of our everyday (digital) lives. Whether we realize it or not, they are becoming our personal assistants, customer care agents, support in finding parking, coffee shops.
Introduction to Chatbot Development
According to the trend forecasters and marketers around the world, last year (2018) was the year of the chatbot. We beg the differ with the attitude – we are living in the chatbot decade and the best is yet to come. To gain better insight into this opinion, let’s take a walk down the chatbot memory lane.
Chatbot development cornerstone was laid in 1966 by Joseph Weizenbaum. The principles used in ELIZA, the first chatbot ever designed, was a foundation for the development structures of chatbots we know today. Although ELIZA was able to fool a couple of users into thinking that they were actually talking to a human, she failed the Turing Test. The test is developed by Alan Turing to test the machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human.
Fast forward to 50 years later and we are witnessing the skyrocketing number of chatbots around the world. Bots have become an integral part of our everyday lives. Whether we realize it or not, they are becoming our personal assistants, customer care agents, support in finding parking, coffee shops and alternative ways to work and sometimes, they just might amuse us like Siri often does.
Picture 1. Short History of Chatbots
What Exactly Is a Chatbot?
A chatbot is essentially artificial intelligence software that can imitate a conversation with a user through messaging applications. It is often described as one of the most advanced interactions between humans and machines. From a technological point, however, a chatbot only embodies the natural evolution of a QA System (Question-Answering system).
It is an assistant that interacts with us through text messages, a virtual associate that can be integrated into your websites, applications or instant messengers. It helps the salesman to get closer to customers. Pal is an excellent example of a chatbot that delivers.
Picture 2. Pal-Chatbot assistant
Why Does a Business Need Chatbots?
Getting rid of routine tasks and simultaneous processing of multiple requests from users are just some of the reasons why. Also, the speed of processing users’ requests with chatbots is tremendous and helps companies in gaining customers’ loyalty. Consumers also benefited from chatbots and they are getting more engaged in this technology.
Main factors that drive people to use chatbots are increased productivity, entertainment, social factors, and curiosity. Chatbots keep the conversions going and enhance social experiences. Conversing with chatbots is also helping in avoiding loneliness, gives an opportunity to chat with someone without being judged, chatbots also improve conversational skills, and all that while they are amusing people by giving them tips, and how to’s.
What are the Types Of Chatbots
Software program for simulating intelligent conversations with humans using rules or artificial intelligence is called a chatbot. Be it written or spoken text, users interact with chatbots in real-time. At the moment we are dealing with three types of chatbots:
Menu or button chatbots
These are the most basic type of chatbots on the market. Most frequently they are decision tree hierarchies presented to users in the form of buttons and menus. Just like automated phone menus in most phone-based customer support, these bots require users to make several selections to dig deeper towards the final answer. Whether the gig-work is a full-time career or an extra income for some, freedom to work whenever and wherever is the hallmark of the gig economy.
These types of chatbots listen to what users are typing (or saying) and respond appropriately, or at least try to. Conceptually, chatbot of this kind follow the customizable keywords to determine how to serve the best possible answer to the end-user.
Chatbots that utilize Machine Learning (ML) and Artificial Intelligence (AI) are known as contextual bots and are by far the most advanced. Using ML and AI, these chatbots store conversations with users to learn and grow over time. Contextual chatbots are smart enough to advance and improve based on what users are asking for and how they are asking it.
How Chatbots Are Changing the User Experience?
Think of all the messaging services people use on a daily basis: Whatsapp, Facebook Messenger, Kik, Slack etc..Count in the equation eMarketer’s survey which found that 63% of customers were more likely to return to a website with live chat and you got the bigger picture – it is essential for online businesses to implement chatbot in order to maximize their performance. Being available for your customers in real-time, when they want it and on their preferred channel, is imperative from the digital, omnichannel society. In the battle for user attention, brands have to stand out in order to take their position on the market. While having a chatbot developed to do so will cut down cost and maximize the time of the customer care agents and improve user experience.
Here is how chatbots change UX for the better:
Improve customer onboarding
Companies that offer products or services that come with a learning curve can drastically improve the onboarding experience for their users. Whenever a user can’t figure out how to use the service, he will have the possibility to communicate with a chatbot and solve the problem quickly.
Picture 3. User experience with a chatbot
Automate transactions and product recommendations
Every business owner wants to make online shopping as convenient as possible for their end-users. Most of the consumers will visit your website with something in mind, but many will not be exactly sure what they want. This is where your chatbot hops in!
Chatbots can nudge the consumer into purchasing by offering them assistance on-site, product recommendations based on previous choices or just by helping them out with the purchase process. In most cases, users will not have to leave the live-chat window to complete an order. Not that different from the in-store experience when you count in handing over your credit card to the cashier, especially if your chatbot has developed a bot persona to make it more human-like experience.
Questions to Ask Yourself Before Developing a Chatbot
This emerging technology is rapidly finding its way to websites of many brands worldwide. With every new chatbot developed, the communication skyline between brands and consumers is changing, but not always for the better. Before getting started with your chatbot project here are 4 questions you should consider:
Picture 4. Questions to Ask Before Chatbot Development
What will the chatbot do?
The role of your chatbot will depend on the customer journey (or a specific part of it) you want to improve. Start by mapping your customer journey and identify places where a chatbot could make the biggest impact.
Many brands have already implemented a chatbot to their customer journey. Some of them allow customers to access information through virtual assistants and messaging apps, others are sales bots serving customers with best-matching items based on the customer’s requests. If you don’t know how chatbots fit in your strategy take a look at some of the best examples from Lyft, Fandango, Spotify and Whole Foods.
Whatever the strategy is, start your chatbot project with a clearly defined set of goals. Keep in mind, chatbot development is a layered project full of opportunities for further improvement and growth. The scenarios, along with the bot personas are fluid as your brand, customers, and trends. In order to stay ahead and fully integrate the bot possibilities in your digital business, you may want to keep track of the bot performance and analytics to detect opportunities for further automation.
Picture 5. Decide on chatbot tasks
Where will the chatbot live?
Where people will interact with your chatbot. Your channel research starting point can be found in your companies data. To ensure the highest possible adoption rate, start with the channel that is already being used by your brand. Your customers are already familiar with those channels and will easily adapt to this upgrade. Some of the potential channels are:
- Your website
- Facebook Messenger
- Your mobile app
Picture 6. Decide where will you chatbot live
What information will it use?
Your bot is smart as the information you provide it with. Whether it collects data from your knowledge base, website, existing documents, shipping information, product inventories, partner sites or some other sources. What information your chatbot will collect depends on the answer to your first question – what do you want the chatbot to do.
Let’s take for an example that you are a wedding dress designer providing people with tailor-made dresses and you want your chatbot to help customers schedule appointments at your wedding salon. To perform the tasks correctly, the bot will need to have access to the next information: Body measurements, Preferred models, Deadline for the dress delivery, Designers’ scheduling and billing system.
In order to map out how the bot will access the data sources, you need to determine what task will it perform. Data collected this way can also be a valuable source for the company in predicting trends and business peak points and job well done in the backend will assure your customers are having the best possible user experience.
Is a chatbot the right way to solve it?
Many of the entrepreneurs are trying to ride the buzz word wave. Before going to the chatbot project, take a second to consider if the chatbot the best way to help your customers. Answering this question will assure that your future customer/users will love what you are doing and that chatbot is not invasive to the user experience.
Also, when talking about user experience, check these UX mistakes that might occur. Knowing these common mistakes can definitely help to design a chatbot that will meet the expectations of your users!
10 Steps to Create Chatbot
So you went through the questions from the previous paragraph and you’ve decided that you need a chatbot. What is the next step? Or should we say steps. In the text below, we will try to further explain the process of developing a chatbot through ten easy steps.
Set goals for your Chatbot
Before you start the development process, identify what your chatbot would do. That includes all required features and implementation tasks. This will allow a chatbot to handle a lot of topics, even include strategies for resolving questions that were normally solved by contact centers.
Your first chatbot shouldn’t be complicated, on the contrary, it should be intuitive and easy to use. Having that in mind, your first chatbot can be a simple “agent like” bot which will route questions to the proper staff depending on the question.
Choose a channel
Text-based chatbots can live on any communication channel that can carry a dialog, whether that’s a traditional mobile channel like SMS, a messaging app, certain social networks, or chat embedded on a website. Whatever channel you prefer, make sure it offers an open API so that bots can be programmatically embedded. The channel you choose can be one you already engage customers on, or it could be completely new to you.
Picture 7. Choose a channel for your chatbot
Develop Conversational Architecture
Chatbots that we mentioned in the first step, the FAQ chatbots, differentiate quite a lot from the ones you find on websites and mobile applications. So in what matter are they different?
The ones you find on the websites/apps have restricted user interactions and have a limited set of options, depending on the user inquiry. Opposed to that, ones that answer FAQs, are open-ended and can say the same thing in different ways. Chatbots are flow-based bots, where interactions that previously happened are always visible to both sides of a conversation. Chatbots also require some kind of a context in the messages so that inquiries can be resolved using information previously collected. Thus, user messages can never be analyzed in isolation, because they are part of a conversation. Therefore, for the first step, you want to create a conversational architecture for your chatbot.
Conversational architecture is similar to an information architecture, which puts the content of a website into a site map. If you’re thinking of developing a chatbot based on a set of responses for FAQs, you should consider responses and which follow-up questions a user might have based on a specific bot reply.
Storyboard and dialog flow design is also very important for a successful chatbot. In the steps of dialog flow design, you will want to decide on what the bot will say at each step. As you will want to design variations of the same message for frequently occurring dialog steps, detailed message design should happen outside of the flow diagrams. This is also known as random prompting, where you can make chatbot use all kinds of variations to essentially say the same thing. This will make customers feel more relaxed because the whole experience will be less robotic and more human-like.
If you have existing self-service platform integrated into some of your systems, same integrations may be also reused for your chatbot purposes. For example, if an application already verified your customer and can provide order information, then it is more than likely the integration can be used for a chatbot. Current technologies, including databases, have connections that can provide a whole lot of detail that is required for a chatbot to function properly. Reaching customer data to respond to customer inquiries is important, but not every chatbot function requires integration.
Collect the data
As we said before, one of the most important resources for a chatbot is a collection of different ways to say the same thing. That’s easier said than done, having in mind the diversity of a human language, and the limitations that are given to a bot by the developer. So how to collect the data? The ideal way is to take the actual text that occurred in conversations with customers. However, if you end up doing it by hand, be sure to have different people on board. This will allow you to get different types of approaches and be prepared when real communication occurs.
Decide on a platform
When deciding a platform you should also think about the development approach. There are two different approaches to these tasks. First one is based on creating rules from the top down, and the other one, which is based on using machine learning algorithms to learn the task from a large collection of previous interactions.
If you don’t poses a collection of interactions, you’ll have to write distinct rules to extract the core meaning from messages. The easiest way of doing so is to find certain keywords in the customer’s message and act upon them. However, that bears the risk in that messages don’t necessarily appear in the form expected. On the other hand, if you do have a large collection of interactions, you could easily apply machine learning algorithms to learn the most common answers to the most common questions.
Implement the Dialogue Flow
In this step, we sum it all together. The conversational architecture, the dialogue flow and storyboard, the platform you have selected, and the data that’s been collected. The main objective of these is the creation of a classifier that will map out the system response according to the incoming text. No matter what solution do you use, it is important to have a diverse set of examples which are as close to the real user as possible.
This step includes testing of a chatbot. It can be done by automated tests and through real-life tests with real people. Testing should be done when the flow forks as it should and you’ve reached the acceptable level of accuracy. Any test that’s done without real user will not give you an honest insight, so whenever you have a chance to test on real-life clients, do so.
Deployment and Revision
So you deployed a bot and work is done? If your answer is “yes”, you’re wrong. This is where work just starts. First interactions should be closely monitored. They will give insightful information on chatbots work, and show a possible space for further development. There may be even more parts that need adjustments, so be ready to work.
Use Cases of Chatbots in Industries
Retail & Ecommerce
Why would you go into the store when you can shop via chat right? Some of the best examples are provided by best e-commerce sites in the world such as eBay, 1–800-Flowers, Nordstrom. May’s OnCall connects will even connect you with a shopping assistant in store in case you need an extra hand in your chat shopping. Here is a list of 6 bots that stand out of the crowd:
In highly competitive travel industry, it is hard to stand out of the crowd. Dozens of websites offer multiple booking options, so it’s not easy to draw the attention of potential customers. Visitors entering and then quickly leaving a website is a common problem in this industry. It’s even more problematic to draw their attention long enough to make any kind of offerer.
Here is a list of 12 bots that stand out of the crowd:
- Great Western Railway
- The cosmopolitan
- British Airways
Picture 8. Travel chatbot
Remember when you needed to make a phone call to hail a taxi? Now, when you text your friend via Messenger that you are on your way, Uber and Lyft chatbots pop up to make your ride as soon as possible. Hyundai and Ford have taken the chatbot game to the next level by allowing users to control their cars via Amazon Echo.
Here is a list of 5 bots that stand out of the crowd:
Beauty and fashion
Fashion and beauty industries have developed sophisticated shopping experiences and a lot of added content for their customers. Facebook Messenger and Kik allow you not only to explore products in chat but to book an in-person appointment with makeup artists (Sephora on Messenger) and recommend tutorial videos based on your previous shopping preferences (BeautyTube on Kik). Personalized, in-app shopping experiences have proven to be the right choice for brands such as Victoria’s Secret, H&M, Aerie.
Here is a list of 9 bots that stand out of the crowd:
- Victoria’s Secret
- Tommy Hilfiger
- British Vogue
- Estee Lauder
Chatbots are the way ahead. They dramatically change the way businesses conduct their sales and customer experience. Although still young, with a development of AI and machine learning, they could be of great use to all of us. Many companies have already recognized the potential of chatbots. Retail, e-commerce, and travel are just some of the fields where chatbots are making their breakthrough.
Chatbots have the capacity to reach out to a large number of customers, more effective than human agents. With further development of machine learning, we may not be able to see the difference of messages we get from a Chabot, from those we get from real-life agents.
Small Business Guide to Amazing Customer Support
More from our blog
What is SaaS software, what does it have to do with clouds and how is it changing the business sphere nowadays? Take a look.
What do customer satisfaction and live chat implementation have in common? Take a look.