How to design the perfect chatbot
Since chatbot-based marketing and customer services are heating up, it is imperative that you develop a chatbot that makes you rise above the pack.
A chatbot is an AI-driven computer program that holds textual or auditory conversations in a usual chat stream on messaging apps such as Facebook Messenger, WhatsApp, or other independent mobile applications for your smartphone. According to a report, 35 percent of customers want more companies to use chatbots for communication and 21 percent of customers see chatbots as the easiest way to contact a business. The bottom line is that your customers expect you to set up chatbots and you need to give them what they want.
Since chatbot-based marketing and customer services are heating up, it is imperative that you develop a chatbot that makes you rise above the pack. It can’t be anything less than perfect. Unfortunately, most companies commit the following mistakes:
Chatbots mimic human interactions. Since they are just programs, they do not have an inherent personality of their own. Hence, conversations with a chatbot, even if they seem almost human, lack a certain degree of character. If you fail to give a ‘botsanality’ to your chatbot, you cannot expect it to perform like humans.
Limited linguistic know-how
Chatbots that use only emojis, other expressive symbols, or limited phrases can’t make logical and personalised conversations. Is your chatbot smart enough to respond even to sarcasm and insults? A few bots don't even know how to land at the right answer due to poor comprehension skills and a frail capacity to separate sentences that don't agree with the pre-built bot library.
When a human customer service executive asks, “How can I help you?” it sounds nice and polite. If you design your chatbot on similar lines, it could end the engagement right there. A chatbot needs to be direct and present a call-to-action immediately. For instance, “Hi, glad to serve you. What kind of pizza would you like to order?” is more direct. Unfortunately, most chatbots stop at the opening question.
After all, a chatbot is a machine and uses your data to communicate. If your CRM framework, classified servers, or other platforms where the chatbot is hosted do not take into account data and system security, a disruption can pulverise everything.
A chatbot should provide information not just with the aim of fulfilling a function but also to create an experience for the customers. So how do you design the perfect chatbot? Here are some best practices.
Give it a personality
Siri has a unique personality. If you tell Siri that you hate her, she will have an uncanny answer for you. Then there is Poncho the Weather Cat which cracks jokes and talks in memes.
Don’t make your chatbot generic. Decide how it should work — should it be a soft-spoken, helpful assistant or a quick-witted one? Let it have a lucid and logical personality. It should be able to reply to a generic “How do you do?” as well as a complicated question like “What time is my flight and what food will be served on it?” equally well.
To form such a dynamic intelligence within the program, it is imperative to keep on testing and refining it. A human-like approach would entice users to interact more with the chatbot and thus deliver a more satisfying user experience. A word of caution here — don’t give the bot your personality! Humans are flawed characters, with differing motivations, tolerance levels, and reactions to situations. You are no different. A chatbot should be an engaging conversationalist without the flaw of being frustrating or irritating to your end clientele.
Now where do you get this personality from? Bot designers usually refer to the ‘Big Five’ personality trait model according to which each personality has five dimensions — agreeableness, conscientiousness, neuroticism, extraversion, and openness. You can take these traits into consideration and design conversation around them.
Build the right conversation flow
One of the most difficult parts of outlining design for a chatbot is to influence the conversation to flow as normally and productively as can be allowed. Most people are new to communicating with chatbots and require some initial direction, and first association is vital to building up how the client will decipher the chatbot. Your chatbot needs to take the initiative. The conversation should ideally have these three steps:
- Introduce itself
- Outline the fundamental services your brand offers
- Suggest the next course of action
Use words, but don’t make it wordy
You want the chatbot to be conversational, not too chatty. A crowded conversational interface will lead to unclear communication. So, it is advisable to design a concise and comprehensible interface. A message comprising 50–200 characters is considered ideal. At the same time, design the chatbot such that it keeps the conversation going and engages customers. It need not talk in paragraphs, but at least prompt the customers when conversations appear to hang.
Take for example a chatbot called Right Click, which helps create websites. It starts with general short questions during the conversation like “What industry you belong to?” and “Why do you want to make a website?” and then moves on to specific questions. Based on the answers, it recommends customised templates. Even if the customers try to digress, it pulls them back into the actual conversation.
Concentrate on the microcopy
Don’t assume that your customer is a highly literate person or belonging to any one gender. Keep in mind who your audience is. A chatbot should be designed in a way that it can cater to them smartly and give clever answers to make discussions more engaging. It shouldn’t utilise jargon or complex grammar and dialect structures. Simplicity is the key.
Protect for user privacy
The security of your users is paramount. If you have deployed a transactional bot that accepts payments or seeks personal information, you need to be especially careful. Use security tools and payment gateways that are reliable. According to a wonderful article authored by Rob Guilfoyle, Co-founder of Abe AI, chatbots can be secured using the following measures:
- User ID authentication
- Two-factor authentication
- Authentication time-outs
- End-to-end encryption
- Biometric authentication
- Self-destruction messages
Planning chatbots requires an absolute grasp of technology and human behaviour. In the event that you consider a chatbot to be another touch point in your brand communication, customer experience and security should be your utmost priorities.
(Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the views of YourStory.)