Rachel Batish's 3 tips to build your own interactive conversational app

In this article, we will provide 3 tips for making an interactive conversational application using current chat and voice examples.

This is an excerpt from the book Voicebot and Chatbot Design written by Rachel Batish. In this book, the author shares her insights into cutting-edge voice-bot and chatbot technologies

Help your users ask the right questions

Although this sounds obvious, it is actually crucial to the success of your chatbot or voice-bot. I learned this when I initially set up my Amazon Echo device at home. Using a complementary mobile app, I was directed to ask Alexa specific questions, to which she had good answers to, such as “Alexa, what is the time?” or “Alexa, what is the weather today?” I immediately received correct answers and therefore wasn’t discouraged by a default response saying, “Sorry, I don’t have an answer to that question.”

By providing the user with successful experience, we are encouraging them to trust the system and to understand that, although it has its limitations, it is really good in some specific details. Obviously, this isn’t enough because as time passes, Alexa (and Google) continues to evolve and continues to expand its support and capabilities, both internally and by leveraging third parties. To solve this discovery problem, some solutions, like Amazon Alexa and Google Home, send a weekly newsletter with the highlights of their latest capabilities. In the email below, Amazon Alexa is providing a list of questions that I should ask Alexa in my next interaction with it, exposing me to new functionalities like donation.



From the Amazon Alexa weekly emails “What’s new with Alexa?”

On the Google Home/Assistant, Google has also chosen topics that it recommends its users to interact with. Here, as well, the end user is exposed to new offerings/capabilities/knowledge bases, that may give them the trust needed to ask similar questions on other topics.

From the Google Home newsletter

Other chat and voice providers can also take advantage of this email communication idea to encourage their users to further interact with their chatbots or voice-bots and expose them to new capabilities. The simplest way of encouraging usage is by adding a dynamic ‘welcoming’ message to the chat voice applications, that includes new features that are enabled. Capital One, for example, updates this information every now and then, exposing its users to new functionalities. On Alexa, it sounds like this: “Welcome to Capital One. You can ask me for things like account balance and recent transactions.”

Another way to do this – especially if you are reaching out to a random group of people – is to initiate discovery during the interaction with the user (I call this contextual discovery). For example, a banking chatbot offers information on account balances. Imagine that the user asks, “What’s my account balance?”

The system gives its response: “Your checking account balance is $5,000 USD.” The bank has recently activated the option to transfer money between accounts. To expose this information to its users, it leverages the bot to prompt a rational suggestion to the user and say, “Did you know you can now transfer money between accounts? Would you like me to transfer $1,000 to your savings account?”

As you can see, the discovery process was done in context with the user’s actions. Not only does the user know that he/she can now transfer money between two accounts, but they can also experience it immediately, within the relevant context.

To sum up tip #1, by finding the direct path to initial success, your users will be encouraged to further explore and discover your automated solutions and will not fall back to other channels. The challenge is, of course, to continuously expose users to new functionalities, made available on your chatbots and voice-bots, preferably in a contextual manner.

Give your bot a ‘personality’, but don’t pretend it’s a human

Your bot, just like any digital solution you provide today, should have a personality that makes sense for your brand. It can be visual, but it can also be enabled over voice. Whether it is a character you use for your brand or something created for your bot, personality is more than just the bot’s icon. It’s the language that it ‘speaks’, the type of interaction that it has and the environment it creates.

In any case, don’t try to pretend that your bot is a human talking with your clients. People tend to ask the bot questions like “are you a bot?” and sometimes even try to make it fail by asking questions that are not related to the conversation (like asking how much 30*4,000 is or what the bot thinks of *a specific event*). Let your users know that it’s a bot that they are talking to and that it’s here to help. This way, the user has no incentive to intentionally trip up the bot.

ICS.ai have created many custom bots for some of the leading UK public sector organisations like county councils, local governments and healthcare trusts. Their conversational AI chat bots are custom designed by name, appearance and language according to customer needs.

Chatbot examples

Below are a few examples of chatbots with matching personalities.

Expand your vocabulary with a word a day (Wordsworth)

The Wordsworth bot has a personality of an owl (something clever), which fits very well with the purpose of the bot: to enrich the user’s vocabulary. However, we can see that this bot has more than just an owl as its ‘presenter’, pay attention to the language and word games and even the joke at the end. Jokes are a great way to deliver personality. From these two screenshots only, we can easily capture a specific image of this bot, what it represents and what it’s here to do.

DIY-Crafts-Handmade FB Messenger bot

The DIY-Crafts-Handmade bot has a different personality, which signals something light and fun. The language used is much more conversational (and less didactic) and there’s a lot of usage of icons and emojis. It’s clear that this bot was created for girls/women and offers the end user a close ‘friend’ to help them maximize the time they spend at home with the kids or just start some DIY projects.

Voicebot examples

One of the limitations around today’s voice-enabled devices is the voice itself. Whereas Google and Siri do offer a couple of voices to choose from, Alexa is limited to only one voice and it’s very difficult to create that personality that we are looking for. While this problem probably will be solved in the future, as technology improves, I find insurance company GEICO’s creativity around that very inspiring. In its effort to keep Gecko’s unique voice and personality, GEICO has incorporated multiple MP3 files with a recording of Gecko’s personalized voice.

https://www.youtube.com/watch?v=11qo9a1lgBE

GEICO has been investing for years in Gecko’s personalization. Gecko is very familiar from TV and radio advertisements, so when a customer activates the Alexa app or Google Action, they know they are in the right place. To make this successful, GEICO incorporated Gecko’s voice into various (non-dynamic) messages and greetings.

It also handled the transition back to the device’s generic voice very nicely; after Gecko has greeted the user and provided information on what they can do, it hands it back to Alexa with every question from the user by saying, “My friend here can help you with that.” This is a great example of a cross-channel brand personality that comes to life also on automated solutions such as chatbots and voice-bots.

Build an omnichannel solution – find your tool

Think less on the design side and more on the strategic side, remember that new devices are not replacing old devices; they are only adding to the big basket of channels that you must support. Users today are looking for different services anywhere and anytime. Providing a similar level of service on all the different channels is not an easy task, but it will play a big part in the success of your application. There are different reasons for this. For instance, you might see a spike in requests coming from home devices such as Amazon Echo and Google Home during the early morning and late at night. However, during the day you will receive more activities from FB Messenger or your intelligent assistant.

Different age groups also consume products from different channels and, of course, geography impacts as well. Providing cross-channel/omnichannel support doesn’t mean providing different experiences or capabilities. However, it does mean that you need to make that extra effort to identify the added value of each solution, in order to provide a premium, or at least the most advanced, experience on each channel.

Building an omnichannel solution for voice and chat

Obviously, there are differences between a chatbot and a voice-bot interaction; we talk differently to how we write and we can express ourselves with emojis while transferring our feelings with voice is still impossible. There are even differences between various voice-enabled devices, like Amazon Alexa and Google Assistant/Home and, of course, Apple’s HomePod. There are technical differences but also behavioral ones. The HomePod offers a set of limited use cases that businesses can connect with, whereas Amazon Alexa and Google Home let us create our own use cases freely. In fact, there are differences between various Amazon Echo devices, like the Alexa Show that offers a complimentary screen and the Echo Dot that lacks in screen and sound in comparison.

There are some developer tools today that offer multi-channel integration to some devices and channels. They are highly recommended from a short and long-term perspective. Those platforms let bot designers and bot builders focus on the business logic and structure of their bots, while all the integration efforts are taken care of automatically. Some of those platforms focus on chat and some of them on voice. A few tools offer a bridge between all the automated channels or devices. Among those platforms, you can find Conversation.one (disclaimer: I’m one of the founders), Dexter and Jovo.

With all that in mind, it is clear that developing a good conversational application is not an easy task. Developers must prove profound knowledge of machine learning, voice recognition, and natural language processing. In addition to that, it requires highly sophisticated and rare skills, that are extremely dynamic and flexible. In such a high-risk environment, where today’s top trends can skyrocket in days or simply be crushed in just a few months, any initial investment can be dicey.

To know more trips and tricks to make a successful chatbot or voice-bot, read the book Voicebot and Chatbot Design by Rachel Batish.

Creating a chatbot to assist in network operations [Tutorial]

Building your first chatbot using Chatfuel with no code [Tutorial]

Conversational AI in 2018: An arms race of new products, acquisitions, and more