Why did people get excited about chatbots?
Facebook debuted its Messenger bot platform in April 2016, and Google announced its Google Assistant in May 2016. During this time, I was on the team that launched The Wall Street Journal bot, which partnered with Facebook, and I witnessed how people’s expectations got so inflated.
Unlike regular mobile apps, users can interact with these bots in chat interfaces. It is an easy-to-understand metaphor for how AI (artificial intelligence) is becoming so human-like. People imagined themselves interacting with these bots just as if the bots were their friends and families.
Chatbots are not a new concept; people had already created Twitter bots for ages. People loved talking about fun bots in their early days, though these bots are seen much more negatively today, associated with being the noise of the internet or consisting of fake accounts pretending to be humans.
In this regard, 2016 was special. Large platformers like Facebook and Google formally announced that they will support business owners to build chatbots on their platforms. More importantly, they presented promising use cases that are natural to peoples’ everyday lives.
These are key reasons for 2016 hype
People don’t need to download a separate app
This is the biggest advantage of building chatbots on these large platforms. Conversations with bots can take place on their widely adopted messaging apps. Service providers don’t need to make an app or website on their own or separately. Not needing to download any additional apps from the App Store is important; If this wasn’t the case, huge marketing efforts would be required to get users to install these apps, alongside product efforts to keep users coming back to the app.
Chatbots can be a simple and cheap option
Frameworks provided by larger platformers can help service providers build basic bots with much less effort than would otherwise have been necessary to build conventional apps or websites from scratch.
Chatbots on these platforms can provide better access to NPL (Natural processing languages)
NYU hosted a consumer study on IVR (Interactive voice response) systems back in 2011. The study showed that 83% of people didn’t find IVR systems useful, unless they were able to understand and respond to straight-forward questions (Consumer Study Finds Overwhelming Dissatisfaction with IVR). Chatbots without NPL may be as useless as these legacy IVR systems. However, by running bots on tech giants’ frameworks, they have better access to advanced AIs, addressing this issue.
Message UI is optimal for mobile experience
Some of the first display interactions on mobile devices were short text messages. Since then, chat UIs evolved to naturally fit into users’ mobile experience.
An example of this is that the chat UI is great for going back through one’s conversation history. You just need to swipe up on your mobile device to revisit content of a recent message.
The trend is still promising
Together with voice interactions such as Echo, Google Home and Siri, the growth of conversational interactions has suggested great prospects for this trend. For example, Amazon announced that more than 100 million Echo and other Alexa pre-installed devices were sold by 2018. Additionally, eMarketer, a market research company, previously estimated that there will be 17.2 million buyers using smart speakers in US by 2018 , and 25.6 million by 2020.
As stated previously, the growth of both chat and voice technology depends on the improvements of NPL. Moreover, the growth of voice and chat bots do and will continue to complement each other.
Today, MAUs (Monthly Active Users) of major messaging apps surpassed MAUs of some of major social media apps. For instance, the growth of WhatsApp and Facebook Messenger has been steady, while that of Snapchat and Twitter is hitting a plateau. The more frequently users open messaging apps, the more opportunities there are for users to engage with chatbots. The growth of chatbots has been predicted to grow at a pace similar to that of websites in the late 1990s.
Now, I would like to address pros and cons for chat UIs in more details, and I hope to reflect these findings for future projects to come.
Not a quick browsing option for nested items
In comparison to mobile apps with rich graphical interfaces, chat UIs require users to take more steps to find what they want.
The retail brand, Spring, was one of the initiators on the messenger platform with a well-executed bot. Their bot showed the world the potential for a conversational shopping experience inside a chat UI. There, users were asked a series of questions such as gender, product types, styles, etc. to find the product they want. However, because retail categories are nested deeply, users had to answer many questions in order to see even a preliminary product list.
Additionally, it is not intuitive to start a new search while users are already deeply enthralled in an existing conversation. Although Spring’s chatbot was brilliant, browsing nested items could be done much quicker in a regular app or website.
Rich UI to supplement for interaction pain points
As stated, chat UIs are limited in their ability to provide intuitive interaction options, which other websites or apps provide. To address this issue, Facebook and Google have provided richer UI options as well as intuitive access to web-view in order to supplement interactions only by text.
For example, Facebook introduced the 2.1 platform with a rich payment interface in July 2017. A new SDK (Software Development Kit) enables users to make payments via a one-step process in Messenger’s webview. This feature can also show payment details that are hard to keep compact without this UI format.
Quick and Smart Reply
In regular chat styles, if users have to type in a message to reply to a chat, they are required to use additional brain resources to keep an interaction going than if they weren’t required to do so. Additionally, free text inputs are often in too random of formats for the system to understand correctly. For example, I often don’t get my desired results when I try to search for a video on my Google Home only with my voice. To address this issue, Facebook and Google introduced Quick and Smart Reply respectively. The quick or smart reply is a list of suggested responses to allow conversations to flow smoothly.
The smart reply is adopted widely into a variety of Google chat base products. In Gmail, users can start writing a response by tapping one of the smart reply options. The name “smart” suggests Google’s seeming emphasis on anticipating responses “smartly”, based on user behaviors. Contrastingly, the name “quick” reply by Facebook, implies that their options are pre-sets and meant for quicker responses. Additionally, these UI elements give unique advantages to a visual chat UI in contrast to voice chat UIs like Alexa or Google Home.
Seamless Sharing Experience
Bot applications can exist alongside users’ everyday conversations. In fact, this provides one of the strongest advantages for bots that run on the messaging platforms. These platforms put immense effort into making user sharing to friends and family as seamless as possible.
To this regard, users can directly access various bots from regular conversations with friends in Facebook Messenger, for example. Users need to simply tap a plus icon which opens a drawer with a list of suggested bots. Once they select a bot, they will have access to unique functions that the bot provides, which enriches their conversation with their friend.
As shown in the screens below, OpenTable, for example, lets users reserve a table at a restaurant and share the information within a conversation with their friend.
Insert Bots based on the Conversation Context
It is even more powerful when bots are suggested effortlessly within regular conversations with others. For example, the example below shows how an Uber ride can be suggested simultaneously as users talk about a given location on Facebook Messenger.
Alternatively, Google Allo lets users add a Google Assistant in a regular conversation. For instance, the example below shows how Google Assistant can insert information relevant to a conversation; a list of restaurants can be suggested if users are talking about a dinner plan per se.
Customer Service Bots
Ultimately, chat UIs are most suited for when a conversation is the most comfortable format for users to interact in. In a survey by Moxie Software, 62% of customers expect live chat on their mobile devices when going on websites, and if available, 82% would use it (Source: ICMI). Evidently, quick and articulate help could drastically improve overall customer satisfaction.
The average customer service call makes people wait, on hold, for quite a long time. The chat as an alternative, could at least suggest answers to frequently asked questions instantaneously — in fact, according to IBM, up to 80% of routinely asked questions can be automated. Additionally, a chat on a display is less stressful than having to wait for a long time to get a response. Moreover, according to Harris Research (ICMI), 53% of customers prefer interacting with an online chat rather than calling for support.
Today, many customer service chats are a hybrid of automated responses and human-operations. This hybrid solution is most optimal for a better customer experience with reasonable operation cost. Users want quick responses and 80% of routinely asked questions may be solved instantaneously with automated responses. For the rest, the system can switch to human operators to provide more sincere support.
Concierge Service for Loyal Users
A concierge service is another unique opportunity that bots have. Kik, a messaging platform from Canada, released bots under their fashion categories, featuring key brands such as SEPHORA, H&M, and Victoria’s Secret.
The new proposition is for users to have one-on-one interactions with these bots. In addition, users can invite friends to chat conversations to ask for friends’ advice, for example. Also, retailers can learn about customers’ preferences through their interactions and personalization options. The service provides a great monetizing opportunity for loyal customers purchasing items directly on the platform.
However, most bots today are still too premature to take full advantage of chat UIs that are well-evolved for human conversations. The premature bots that only routinely update content wouldn’t make the interactions as vibrant as the conversations with your friends. For example, I find it more useful if the concierge bot is not use by each brand, but rather having one concierge bot that suggests products from a variety of brands.
All in all, the role of a given platform is extremely imperative to pushing this bot trend to the next level. In early 2018, Facebook discontinued its virtual assistant and M. Google discontinued Allo as well. This news implies how these bots are not mature yet and are not straightforward enough for people to talk to directly. Voice has a unique advantage, because it’s completely hands-free. I am looking forward to applying this knowledge to future applications and seeing how these bots will become more practical and useful as time goes on.