Designing both didactic and effective chatbots means having a very clear idea of what your virtual agent will tell your customers about your organization. To arrive at this clear vision, it takes time, patience and an understanding of several key points related to the bot.
The realization will also require a close working relationship with a software developer and a long-term follow-up. But the game is worth the effort.
The error is human
"The bot is a new medium. Solutions [to make a good bot] are not obvious or immediate, "warns Beerud Sheth, CEO and co-founder of the Gupshup Chatbot platform . "You will have to experiment and therefore make mistakes. "
No organization has yet deployed a perfect chatbot. Like humans, chatbots can misinterpret words - and make mistakes.
Preparing for these unavoidable mistakes is one of the first things your organization needs to understand when working with a virtual assistant developer.
Identify the purpose of having a bot
Do not expect a chatbot to provide much value if its only use for you is to get noticed.
On the contrary, have an operational objective and a precise idea of how it will be used and why. Plan its role as you would for human employees.
Most Gupshup customers, for example, want their chatbots to help increase the number of their customers. The second goal is to improve the amount of sales per customer, and then support assistance.
Once you have defined the chatbot's tasks, you will need to select a team that will work on these issues with the developer - from design to implementation, to managing and tracking the daily use of the chatbot.
"We do not need technical or linguistic experts on this team. Instead, you need people with in-depth knowledge of the business and trades, "says Mariana Martinez, a former content and marketing specialist for the Aivo Virtual Assistant platform .
A business-oriented team will indeed be better able to answer the critical questions that will make it possible to adapt the chatbot to its audience.
How to customize your bot
Most companies already have an image, a color, a logo and sometimes even an avatar in mind when designing chatbots, says Martina Martinez. For her, design is indeed important, but tone, language and vocabulary are as important as they also shape the personality of the bot.
This personality must match the image of the company. Gupshup, for example, has created a Chatbot for Skittles Candy that mimics the brand's colorful and fun marketing messages. "It takes a lot of brainstorming to do that. It is essential that the personality of the company express itself. The owner of the brand must be involved in the whole process, "says Beerud Sheth.
The case All About Jazz
The character of the chat chat Facebook Messenger of the site All About Jazz tries for example to adapt to the wide range of its readership.
Neophytes type basic jazz questions - information about an artist or a concert date - while confirmed hobbyists ask questions about the composition of a group on a particular recording or about the notes of the group. a cover of a rare album.
The secret is to make the chatbot's personality incarnate in the language, with words and phrasing that bring them closer to the users.
The personality of the chatbot is also found in his name. Michael Ricci, founder and publisher of All About Jazz, has chosen "Bix" as a tribute to 1920s jazz cornetist Bix Beiderbecke. Old school jazz fans will make the link, while "neophytes will think it's just a cool name," jokes Michael Ricci.
The choice of words (written or spoken)
The words a chatbot uses to communicate - and the words that it will be "trained" to understand - will be methodically determined by the client and designer before being implemented in the software. This is perhaps the most difficult and longest development phase. After all, a chatbot is supposed to "chat", either verbally through a speech recognition platform and language generation or through typed messages via textual channels or social media.
Structured Conversation vs. Unstructured Conversation
A chatbot's conversation can be conducted in two ways: structured and unstructured.
In a structured conversation, the chatbot puts barriers around the conversation by using selected terms so that the human, stuck with limited choices, will not derive any.
This type of chatbot can even display link responses to bring web and mobile visitors directly to what they want or (should) see.
"When you do a structured conversation, there is no chance of misunderstanding," says Beerud Sheth. A structured conversation, he adds, can be adapted in more than 200 languages "very easily".
For a more talkative chatbot, the answers will come from a repository of formulations that are assigned, upstream, to the words used by customers. Remember that it takes years for humans to understand each other almost completely. And this is not always guaranteed because of a misplaced word or an unusual idiom.
Bots are like toddlers and are just beginning to "understand" the sentence structure.
"It's a lot harder when the conversation is not structured," says Beerud Sheth who recommends carefully choosing all the words that the chatbot will use through a structured language format and, in a second step, to give it more to handle gradually.
Artificial Solutions also warns customers not to have too high expectations in this early age of chatbots. "The problem is they're pretty stupid," he laughs. "They follow a linear path and can not deviate from it".
Accompany the baby bot
When designing chatbots, do not leave them alone when they do not understand certain human voice commands. With semantic and voice recognition technologies, chatbots can learn by listening to words and phrases that are not in their repertoire and gain experience. Provided to accompany them.
Starting a bot project involves following the process over time.
"Even if everything works well, customers may not be satisfied with the answers given," says Artificial Solutions' Andy Peart. "Businesses can and must exploit these returns. With this data, they will see when people are unhappy and what is causing it. You will be able to refine your bot ".
An increasingly natural technology for the user
In any case, the bots seem to be here to stay. Their uses are multiple: help call centers, automate basic responses (and increase engagement), optimize document searches, support the use of tools. And other opportunities will certainly follow.
Gartner estimates that within two years, 25 percent of customer service and support operations will integrate at least one virtual assistant technology across all of their engagement channels.
"We did surveys on voice assistants. Adoption is growing. People expect more and more to be able to talk to devices like a human. Soon, every company will be expected to have one. "
0 Comments
Post a Comment