How to create a Chabot

What is a chatbot?

A chatbot is a computer program that is capable of having a conversation with a human being or with another chatbot.

We can use chatbots on our website, in an app, email replies, SMS conversations or in communication and messaging tools such as Facebook Messenger, Telegram, WeChat, Line, Kik, Viber or Slack.

Related topic : How to connect your phone to a camera?

The uses of chatbots are diverse, as we will see in more detail in this guide, but for example we can develop chatbots capable of providing customer information about our terms, managing orders, making reservations or handling incidents.

What types of chatbots are there?

According to the intelligence of the chatbot

At this point we need to introduce two concepts: NLP and Machine Learning.

You may like : How to use a chatbot?

  • NLP (Natural language processing) is a branch of artificial intelligence that would make it possible to create a chatbot capable of understanding what a person who expresses himself in natural language (i.e., as would two humans communicate) and, on that basis, be able to formulate a answer. An NLP system may be manually programmed (with a complex set of rules created by hand) but its true power starts when we introduce Machine Learning.
  • Machine Learning is another branch of artificial intelligence consisting of the development of techniques that allow machines to learn autonomously and automatically.

Among the AI tools applied to the world of chatbots, possibly the most popular is Watson from IBM, but there are other alternatives such as Lex from Amazon, Luis de Microsoft, Facebook wit.ai, Google api.ai, or Cisco MindMeID.

Is artificial intelligence essential to create a chatbot? No, you don’t need to have any NLP or machine learning systems to create a chatbot.

We have other alternatives to create less complex chatbots. Our chatbot can present various response options to the user in the form of a dialogue so that the user selects one of them and, based on the choices, our chabtot will give answers. In other words, it is about guiding the conversation, having previously planned the possible answers or needs of the user, so that, since they are closed answers, there is no need to have an intelligence capable of interpreting the user’s words.

We will see examples of the latter later. As an advance, in the next screenshot we see the configuration of a chatbot conversation tree made with the Motion AI tool. We look at the different paths that the conversation can follow. Each of the white boxes would be each of the questions/actions of the chatbot, with its own configuration, and which, depending on the user’s answers, would lead to one step or another.

Setting up a chatbot conversation tree with the Motion AI tool According to the chatbot development platform

We have to differentiate between the intelligence that makes the robot communicate and the platform on which we develop the chatbot, which would be the technology or program we use to develop and configure our chatbot. In practice, both concepts can be merged, but they are different issues.

We should not confuse this development platform with the environment where you will later live our chatbot and have your conversations, such as a messaging application. Although in practice it may happen that the environment where we want to host our chatbot offers us a development platform for the chatbot.

To the point: how can we develop a chatbot?

There are several alternatives. The first of these, as we mentioned, are the official frameworks . Many messaging and communication apps provide free of charge a set of libraries and resources to help developers create chatbots that will work on their channels. Such is the case with Facebook Messenger, Slack, Telegram or Kik. Microsoft also has its own framework that allows you to develop chatbots that can work on all major channels.

On the other hand, we can resort to Unofficial frameworks such as Botkit, ChatScript, Meya, Gupshup or Pandorabots. In this case, some solutions are free but others are paid.

In case we don’t want/can resort to programming, there are trading platforms that allow us to create our chatbot from scratch at really cheap prices. On many of these platforms, even if we don’t have the need to throw code, there is the possibility, if we want, that certain parts of the chatbot develop them programmatically.

Examples of trading platforms include Botsify, Flow XO, Rebotify, Morph.ai, Motion AI or Smooch. It is also worth mentioning a few specific to chatbots on Facebook Messenger, such as Chatfuel or ManyChat.

Finally, we have another possibility: buy a predefined chatbot template that we can later customize. We can find shanty template stores in BotMakers or the Motion AI Bot Store.

Depending on the option chosen, the development platform may not provide us with how to integrate our chatbot with the environment where we want it to work, such as our website or a messaging app. We may also be easier to integrate but by paying either a fixed fee or a variable per use. To avoid either of these two cases, there are tools that integrate our chatbot wherever we want it to live. This is the case with Recast.ai Bot Connector, with wide compatibility, as well as being open source and free.

What uses can a chatbot have?

First of all, you have to think that this is not black or white, human or robot. That is, we can have processes in which part of the conversation with a user is in charge of a chatbot and other parts where a human takes over the command of the conversation.

Customer service

To this day, completely replacing customer support with a chatbot does not seem very viable. However, a chatbot is extremely useful as a first point of contact with the customer when they want to make issues such as handling a complaint, checking the status of their order, giving us feedback on the service or consulting our terms and conditions, to name a few examples.

In these cases, it is common for the first customer interactions to be the same in all cases, with questions we repeat to all customers. It is usually a matter of first understanding what the customer needs or wants and, after that, asking them for the information needed to deal with their case or request.

A very basic chatbot can be very efficient by collecting the information we need. And even, depending on the case of the customer, you may be able to answer it. When a human gives over to the robot and takes the case, he won’t have to waste time collecting customer data or giving recurring information to it.

On the other hand, a chatbot is available 24/7 . 24 hours a day, 7 days a week, so it is also very useful as a complement to our customer service when our company does not offer customer service.

Similarly, a chatbot can chat with unlimited people at once , so it can also be an add-on at times when there is an excessive burden of customer service.

Sales channel

This case is similar to the previous one. It will be hard for us to develop a chatbot with the commercial skills of a human, but it will be possible that such chatbot can streamline part of the conversation and channel it.

In commercial work, there are usually a series of repetitive and common questions that aim to understand what type of product the customer needs. A very basic chatbot is able to perform this task, as well as allowing you to display a series of products according to what the customer needs and provide useful information.

An example is the SnapTravel hotel search engine. In addition to other conventional sales channels, SnapTravel allows you to find hotels thanks to a chatbot operating on Facebook Messenger, Slack and via SMS.

SnapTravel Chatbot on Facebook Messenger In the screenshot we see how the chatbot, after asking you a series of questions, gives you results according to your search. It also gives you the ability to further filter the results, do a new search or ask to chat with a human agent. In the same sector, they also offer a similar Skyscanner service for flights and kayaks.

Another example of using salesforce chatbots in an e-commerce was facilitated by David Marcus, vice president of messaging on Facebook, at the launch of the Bots API for Messenger during F8 2016.

We must also think that even if we as a company have clearly defined what our sales channels are, a potential customer may request commercial information by other means. We have to be prepared for it and there a robot can help us when we don’t have the ability to have staff on all those channels.

When considering using a chatbot for this purpose, you have to take into account, of course, the coldness of talking with a robot instead of a human being. Although such coldness is not always negative. Some people find it even more comfortable to interact with machines than with people. Who hasn’t said that “ask no” thing when you have to call the pizzeria? In other words, we can even encounter scenarios where a chatbot is more efficient than a human in commercial work.

In addition to the aforementioned advantages that a chatbot is available 24/7, that it can have unlimited conversations and that it saves time for the person of the company, there are two other commercially important advantages:

  • The chatbot’s response is always immediate . We don’t keep the customer waiting while he sees “writing…”
  • The answer of the chatbot is always correct and always use the most appropriate words from the point of view of our commercial argument.

Informational

Chatbots also provide another way to consume information. An example of this is Techcrunch’s chatbot on Facebook Messenger, developed with Chatfuel. This chatbot allows you to interact to receive news in sections, whether on request or on subscription, automatically receiving news and articles that interest you at the time of day you decide on Messenger.

TechCrunch Chatbot on Facebook Messenger Techcrunch is not the only medium that uses chatbots to share news and send notifications to its readers. So do other media outlets such as Wall Street Journal, Washington Post, Today or BBC.

Another example is Hi Poncho, an app available on Apple Store and Google Play whose utility is to send weather information with a touch of humor. Thanks to its chatbots (available on Facebook Messenger, Kik, Vivber and Slack), it is possible to receive the information without installing any app.

Chatbot Hi Poncho on Facebook Messenger In the case of Johnie Walker, the Whisky brand, takes advantage of the possibilities of chatbots by offering information on Facebook Messenger about their brand, help choosing from its various varieties according to our tastes and budget, cocktail proposals, a whiksy tasting guide and a list of where to buy their products according to your zip code.

Other uses

We can find chatbots that are our personal fitness trainer, that help us send money and receive information about types of changes or speed up collecting answers, such as Howdy, a robot that does typical rounds of questions for you repetitive, it collects the answers and sends you the results when everyone has responded.

Among the professional uses are two chatbots that, for meetings, help to square the agendas: Ami and Meekan. And for rest times at work, games like Emoji Salad, a pictionary with emoticons. The chatbot Emily has a more complicated task: talk about death. Another example: Paramount released a chatbot to promote the Ninja Turtles movie.

And we left out personal assistants like Siri, Alexa or Cortana. But that’s another league…

How much does it cost? Is it expensive to develop a chatbot?

Well, it depends. It depends on the intelligence behind the chatbot, the chatbot’s development platform or computer development needs.

For really complex chatbots that need real intelligence artificial, with a powerful semantic interpretation, then the price is prohibitive. Only available to large companies.

But there are alternatives for every budget since, as we have seen, there are free development platforms that allow us to create modest chatbots but with interesting features. In addition, if we don’t have access to programming resources, several of those platforms allow us to develop chatbots without having to throw a single line of code. To give us an idea of how cheap it can be, and that it is not just a technology available to large companies, for less than €50/month in motion.ai we can develop and maintain up to 25 chatbots with a limit of 20,000 messages per month. And without the need for programming knowledge.

Conclusion

Chatbots as we would like, able to have a smooth conversation with a human and give Coherent answers (and wise, from the point of view of a business), are very expensive and not even entirely reliable. The technology is not developed enough to put a chatbot in charge of important conversations or criticism in our companies.

At least the technology that is available to most companies since, for example, we have recently seen how Facebook has had to disconnect two chatbots because they had developed its own language. It was an experiment in which two chatbots, programmed with artificial intelligence and using machine learning, had to talk to each other. The objective of the test was to study and develop negotiation techniques using this technology.

In the short term, perhaps we shouldn’t be overly demanding on chatbots. This is all really starting. However, in the landscape of chabots, today we have very economical possibilities that allow us to automate certain conversations, with obvious advantages:

  • A chatbot is available 24/7.
  • A chatbot can chat with unlimited people at once.
  • The chatbot’s response is always immediate.
  • The answer of the chatbot is always correct, in the sense that it responds exactly what the company wants it to respond, without misleading or confusion that humans sometimes have.
  • Open another channel for consuming information and subscription by customers.
  • It saves time for staff in charge of repetitive conversations, and also eliminates the frustration often caused by this type of task.

In Ultimately, there is still a lot of potential to develop in this field, but from now on, we can make the most of chatbots and economically.