The art of designing a chatbot experience
Chatbot conversations go beyond a simple yes or no. It involves understanding the complexities of human language and responding in a way that feels human.
So, you want to design your own chatbot.
Great! You’re taking the first step towards automating the repetitive tasks and improving profit margins from your platform.
But where to start? After all, designing chatbots should be easy - just slap some dialogue together and feed it to a bot, it can take it from there. Right?
Wrong!
Human conversation is nuanced, multi-layered and complex. And designing a computer program that can understand our language and provide the help we ask for, requires thoughtfulness and planning. But don’t worry, we’ll equip you with the fundamentals of designing chatbot conversation to help you get started.
Chatbot 101
Any internet user exiting in this millennium would be more than familiar with chatbots. You go to a website and a pleasant ting indicates to you that there is a friendly assistant at hand to answer all your questions. Sometimes they’re helpful and sometimes decidedly not. In most cases however, you get a shortcut to the information you’re seeking on the site or the goal you’re trying to achieve.
But what are chatbots really?
Chatbots are computer programs that can simulate human conversation with the people using it. Most chatbots are designed with a specific goal in mind, and offer very little outside of that. For example, a bot on a sales related site will be focused on generating leads.
Nowadays however, general purpose bots are much more common (see: chatgpt). In fact, with the advent of machine learning, deep learning and AI technologies, modern day chatbots are appearing with the ability to understand and converse organically with humans, and even improve with more exposure.
That is precisely why now more than ever, there is a disruption in the industry thanks to the advent of conversational AI. It is said that chatbots have increased sales by around 67%. (Source: Outgrow). By 2027, it is expected that more than a quarter of the businesses worldwide will turn to chat technologies as the primary communication channel with its customers (source: Gartner). These emerging trends in the chatbot industry are the reason companies are investing more and more into conversational AI technologies.
Besides the tech, conversation design is also garnering more visibility and respect as its own niche field of design. Gone are the days where the makers of a chatbot haphazardly write a few lines and call it a day. Dedicated conversation designers are popping up and are responsible for creating the flow of speech within a virtual assistant. And these flows always start with what is known as Sample Dialogue.
The Building Blocks
Sample dialogue
Sample Dialogue is one of the first things one would plan while designing for chatbot conversation.
When designing Sample Dialogue, you create imagined scenarios of user and bot interaction and write back and forth dialogue. Doing this for as many relevant situations as possible can help you visualize the overarching conversational style of the bot and make corrections to the interactions as you see fit.
Here’s an example
User: Hello
Bot: Hi there! Welcome to Holiday Rentals. How can I help you today?
User: I wish to book a trip to Paris this year
Bot: Great choice! Paris is a beautiful city, with many amazing sights and activities. May I know when you plan to make your trip?
User: December 20th, and stay up until new year
Bot: Good to know. What type of accommodation are you looking for?
User: A private villa would be nice, hopefully pretty close to the main attractions.
Bot: Perfect! What kind of main attractions are you thinking of?
User: The Louvre museum maybe?
Bot: Great choice! Being the largest art museum in the world, the Louvre is home to some of the biggest masterpieces in history and spans roughly 73,000 square meters of exhibition space.
Bot: Do you have any other preferences?
User: No
Bot: Okay then! I have narrowed down your rental options to around 62 properties. Would you like to check them out?
By creating multiple back and forth conversations, and reiterating them, you can successfully design a full fledged conversational bot.
NLP
NLP or Natural Linguistic Programming can help a bot process conversation in a manner that is similar to human understanding.
It can equip the bot to process open-ended or vague statements, and understand the user despite typos or translation issues. They can go as far as to understand what the user’s intent ultimately is and make suggestions accordingly.
NLP is often present in conjunction with Machine Learning and artificial intelligence technologies. It works by converting the human input into a form that is understandable for the bot, and then generates a response accordingly that is understandable for humans. Besides NLP, as a conversation designer, there are certain principles you can follow to keep the interactions efficient and welcoming for the user.
Principles of Chatbot Conversation
The reason chatbots often fail to meet the expectations of users are because they don’t account for natural human tendencies . The tendency to forget the exact name of things. The tendency to not have the right words to express thoughts. And the tendency to not really know what they want.
If you want your users to actually feel seen and understood, it might be a good idea to approach the process with a little more empathy.
- 1. Maintain the 4Cs
- Clarity: Use the simplest possible language that gets the point across. This way you can create a chatbot that makes sense to the entirety of its user base and not just the ones who are familiar with jargon
- Context: Ensure that the context is always clear, i.e. all the dialogue is relevant and draws from the earlier instances of the conversation.
- Consistency: This is a two part principle. One, try to keep the branding or personification of the bot consistent with the business values and tone. The other is to keep the chatbot consistent with itself, especially if it is offered across various channels. For example, a company called Kayak currently offers chatbot services on FB messenger, Slack, Amazon’s Alexa and Google assistant.
- Conclusion
When a user comes to a chatbot with a specific goal in mind, there needs to be a natural stopping point once it has been achieved. This needs to be intimated to the user with some sort of feedback or confirmation, so that they are aware of the end instead of waiting for something to happen.
- 2. No gaslighting
A huge portion of users actually prefer human interaction to chatbots. A survey in 2019 on consumers in the US found that over 86% would prefer to talk to human agents (source: forbes.com). A part of it is trust issues and another part of it is just the deep seated need for human connection. However, most people do deign to use chatbots if the task is trivial or they need something quick. It is important that they are aware of the fact they’re interacting with a bot, because learning it after the fact can make them feel a little cheated.
- 3. Personify your bot
While people do not like being tricked into believing they’re talking to a human, it still is a good idea to give your chatbot some human-like qualities. It would elevate the whole experience from something robotic and dull, to something fun and engaging. In order to do this, usually a UI/UX design agency tries to design a persona for your bot. A chatbot’s persona is usually tied to the business it exists in the context of. Therefore, a real estate company might want a bot that is professional and welcoming, while a business dealing with adventure sports might opt for something more casual and fun.
While planning for the bot’s persona, consider the personality, tone of voice and the way it behaves. Consider giving it a name and an avatar to match.
- 4. Plan for human error
To err is human and in a chatbot interaction typos or misinformed queries are very common input errors. . How the chatbot handles these errors can make or break the interaction. Intimating the user about the error and suggesting ways to course correct would be the best move in this scenario, as opposed to letting the user grow frustrated about not getting what they need.
Conversation design by nature is a vast expanse covered with landmines. Navigating the tricky waters that is human-computer interaction is always a monumental task, but with a few considerations and tricks up your sleeve, you too can design a chatbot experience that is fulfilling and easy to use, and ultimately drives the goals of the business forward.
