The definitive guide to building a chatbot

10 steps to help you control conversational interfaces

Do you think Tinder is much more than a dating app? Read what we tell you here! The 'Swipes' on Tinder to choose or dismiss a person mark a turning point in UI (user interface). A change in the rules of the game on mobile screens. A new interaction that breaks with the established norms between a user and a mobile device. In short, a micro-revolution.

The example to have is a clue to the major changes that UI will soon experience. The emergence of motion detectors, virtual reality, or augmented reality will cause a paradigm shift. But undoubtedly, the indicator that anticipates this UI revolution is the chatbot. This is what is most known because it has been talked about a lot. Today, few deny its potential and transformative capacity when it comes to relating brands and consumers, offering personalized experiences, and scaling as much as desired through artificial intelligence (AI).

The advantage and disadvantage of chatbots is that building them can be as simple or complex as desired. At SEIDOR, after creating chatbots of different sizes, we have acquired knowledge about their conceptualization, definition, prototyping, and programming. And yes, it is possible to create one with the right tools and profiles in a hackathon in just 72 hours. But perhaps it is not the most optimal. To succeed in creating a chatbot, these are the steps!

Set up a transdisciplinary team

A chatbot is not created by a copywriter alone with an editing tool. Its role is as important as that of a UX expert, the support of a UI profile, and of course, a developer who knows the chatbot platform: from a website to Facebook Messenger, Slack, or Whatsapp for Business, among others.

Get to know the brand and its audience

Immersion in your client's digital business is essential to define the chatbot's objectives. They will often explain that they have two objectives: to increase engagement and reduce response time. Go a step (or two) further. Explore, navigate, delve into and understand how social media works, read the FAQs, and make inquiries where you believe a chatbot will be useful to your client. See how the brand interacts with its audience and vice versa. This analysis is crucial to offer a proposal that covers the KPIs and, on the other hand, to create the chatbot from a user-centric vision.

Build a specific chatbot

When building a chatbot, make it specific. It's pointless to create one that does everything despite the temptation to make a multi-purpose bot. A recent study reveals that 70% of the 100,000 bots on Facebook Messenger do not meet users' requests. This is partly due to not knowing how to approach bots from the audience's perspective.

Card Sorting exercises are extraordinary for defining and prioritizing how we will use the chatbot. It is also a first approach to adapting the brand's tone of voice. For example, if we need a chatbot to answer frequently asked questions about a product or service, we will need to know what the 10 most requested questions are, which ones receive the worst response from the user, or which ones are the most difficult to resolve.

Adapt it to the tone of voice

Brands go beyond the logo and naming. They are a set of experiences with multiple points of contact with users. Chatbots are one more. Therefore, the conversational layer of the chatbot must maintain a narrative consistent with the brand story. We achieve this by giving it its own personality and tone of voice. Access the brand style guide before writing words or expressions, co-create with customers to define the bot's personality... You can make it as complex or as simple as you want.

Create a conversational tree

To date, the vast majority of bots have been built following the logic of a decision tree. Very cold, all. That is why it is essential for the UX designer and the copywriter to align user journeys and technological specifications with the conversational tree. They must include all points of contact with users and the bot, anticipate error messages or inappropriate (or directly insulting) reactions from people. You can even identify the moods of users at each of these points of contact to include elements such as emojis or iconography. The decision tree is basic before implementing the tone of voice.

We recommend app.diagrams.net to create collaborative conversational trees. It is a deliverable that should not include the final content but will be very useful to align workflows between the UX team and content creators. Use codes to identify each text in the tree and use them as captions in a collaborative Excel document.

Write and Rewrite

You have all the cards on the table. Now is the time to write. And rewrite. Not only do you have to choose the right words to convey the tone of voice. You must also take into account the structure of the conversation and the user's moods at each touchpoint. Focus on a facilitating language to enhance the usability of the chatbot. In other words, 'full on' UX Writing to achieve coherence and not break the established usability in the conversational tree.

Don't let the word robot influence you! You write for people and "you-don't-want-to-seem-like-a-machine". That's why, write the chatbot content in a conversational language. Think about your WhatsApp conversations. Use emojis and visual material to help the user chat and navigate through this visual interface at the same time.

Rapid prototyping

So far, everything has been happening in an Excel and in conversational trees. Although they are optimal working documents, you will bore your interlocutors when you present them. They will understand better if we put them in the users' place. That's why it is important to quickly prototype the chatbot design. It is essential in the bot creation process.

There are free rapid prototyping tools for chatbots. One of the most recommended, despite not being infallible for all platforms, is Bot Society. When using it, be careful with the character limiter or the use of images.

Choose a reliable platform

Facebook Messenger is the most popular platform for hosting chatbots. But even though it may seem like it, it is not the only one. Choose the one that best suits your client's needs strategically and technically: Slack, Whatsapp, Twitter, Telegram... Maybe a native chatbot? At SEIDOR, through our Venture Builder The Carrot Cake, we have supported the company Hubtype, a great ally when it comes to building and managing bots. This company located in Barcelona won the award for the best Chatbot Startup of 2018.

In addition to bot development, Hubtype allows for efficient bot management, tracking of chatbot usage metrics, and most importantly: absolute control over the conversation with users. The folks at Hubtype have put together a quick guide to create a chatbot on Facebook Messenger in just 3 steps. Their article, combined with our ultimate guide, will be your chatbot bible.

Don't forget the VUI

We don't just write for users. We also speak to users! At SEIDOR, we have developed a concept with Google Assistant to book a meeting room. You can read the case study here. But consider defining a chatbot as the first step towards evolving into a voice digital assistant. If you have defined the tone of voice well, giving it a real intonation and spoken expressions will be much easier.

Test me if you can

Perhaps it's obvious, but we never tire of repeating it. Testing will put the chatbot (and any other service or product) in its rightful place. Before the launch, test your product with as many users as you want. Whether it's in a user lab or a Guerrilla Test. But test! It will help you better understand the audience and shatter hypotheses that you considered absolute truths. Because without a doubt, User Killed The Technology Star!