Chatbots are already ubiquitous, at least in the home and on the phone: a recent report shows that between a quarter and a third of Americans already own a smart speaker with a voice assistant, making it one of the most rapidly adopted new technologies in history.
And usage will grow: 48 percent of consumers are expected to own a smart speaker such as Amazon Echo, Google Home, and Apple HomePod in 2019. When the technology matures to work almost perfectly, which is not yet the case, adoption will not only come close to 100 percent, but each consumer will own multiple devices (including cheap, commoditized devices) that support voice.
Read More: Unattended bots to take on govt jobs without ‘standard work hours’
It’s important to note that while most people’s experience of a chatbot is via the phone or speaker device, the way these chatbots currently work is not how they will work in future, especially in a business context.
Smart assistants targeted at consumers are currently very broad and shallow. They are controlled by one-off commands and questions, which makes them great for some tasks (playing music off Spotify, checking the weather, setting timers, etc.) — but makes them unusable for tackling complicated tasks in the business workplace.
Consumer chatbots are focused on understanding the initial command given to them. There is no “dialog” between the user and chatbot (where the user can respond to what the chatbot said and get further contextual replies), except in rare cases where the interaction is mechanical, such as telling jokes.
However, dialog is central to business bots. Business chatbots focus on more complex tasks and higher value customer jobs, both of which require multiple dialog turns. This is possible because business chatbots focus on a narrower topic domain than consumer bots, which means they can be more accurate within a given context.
This focus on a narrower topic domain is critical given the limitations currently faced in artificial intelligence, specifically with regards to natural language processing and dialog management. This advantage in limiting context, and therefore the potential questions and statements that the bot needs to understand, means that business chatbots can appear to be much smarter and more useful than consumer chatbots.
When it works well, it can seem almost like magic.
The following are (obvious) examples of how a business bot might derive its context from the situation:
Of course, the reason business chatbots are “smarter” is not just that they are more topic-focused, but because more manual development effort goes into developing them. Consumer bots, like Google for search, must almost totally automate building understanding given the scale of topic coverage.
Business chatbots, however, are purpose-built for specific tasks, which means holes in understanding can be plugged manually by designers. This applies to the original setup of the bot and to design modifications made after the bot has gone live and performance analytics are available.
Read More: When chatbots become more than just machines
This does not mean that business bots do not use the latest machine learning techniques to build understanding automatically — they do. What it does mean is that the shortcomings of artificial intelligence (which are very apparent on voice assistants right now), can be overcome through thoughtful design.
Of course, chatbot development platforms focus on using AI to make the work of human designers in building understanding for the bot as easy as possible, but do not aim to eliminate human designers — far from it!
Human designers are needed to craft the seamless experience for end users. It is possible in many cases for business bots to have human support agents as back up to the bots (called Human in the Loop).
If the bot doesn’t understand the question or statement, it can be escalated to a human agent. This is not possible for consumer bots, given their scale. The information that the business bot gets from the human agent can be used to make the business bot smarter (using machine learning), which is a further advantage for business bots.
The better customer experience with business chatbots (and bots in general) does not need to be limited to text or voice. End-users of bots will benefit if the bot uses, or has two-way integration with, graphical UIs such as websites, web forms and web apps.
While human agents are limited in the way they can respond to customers’ text or voice, bots are not limited. If a graphical interface is a better end-user experience, the bot, unlike the agent, can show a graphical UI to the user in the chat window, app or on the website as appropriate.
Expect to see even more integration with screens, beyond the Echo Show and similar devices, with consumer bots as well.
A nice example of how dialog turns work in business chatbots is the Google Duplex demo. Of course, this demo cherry picks the best example out of many attempted calls, but it still gives an idea as to where the technology is heading. In reality, however, what is needed is this sort of end user experience every time.
Many of the technological advances in chatbots apply to both business and consumer bots. All bots need improved NLU, or Natural Language Understanding, the functionality that allows bots to produce meaning and understanding with nearly human-like accuracy.
Future bot and AI research will focus on creating more contextual NLU, better speech recognition, and faster response times to create a smoother, slicker interaction between human and the bot.
Business chatbots will leverage usability research and advances in the consumer chatbot space, but these advances will be applied to much more complex interactions.
Expect to see a much better performance of business chatbots in understanding context than consumer bots.
Voice technology needs to improve. For voice interfaces to work really well, continuous processing of the voice will be necessary instead of relying on hot-words. Instead of having to explicitly reference Alexa in a new statement as you do now, e.g. “Alexa, can you remind me to […] tomorrow at 8 a.m.?”, you will be able to say, “Oh, and Alexa, please remind me that tomorrow at 8 a.m.” in the middle of an ongoing conversation at the workplace.
This will be achieved by businesses processing the voice fully on their premises, or by putting in place an offline voice buffer mechanism for cloud processing without continuously streaming.
There will be cheap commoditized voice devices in the future, so that devices can be put in every room. For businesses, it will be crucial that these devices do not record every single conversation in the office, but use offline technology as appropriate to protect privacy.
The crowdsourced development and sharing of natural language processing data and other necessary AI data will speed the development of smart business bots. Think mechanical turk built for purpose with AI-driven facilitation tools (such as synonym suggestions) thrown in — at least, that is where the technology is going.
Whether it’s as a customer or employee, it will become more and more common in the future for you to interact with business bots to get answers to questions, complete tasks and generally track and get you the information you need.
According to Juniper research, chatbots will be involved in 85 percent of all types of business-customers interaction by 2020. A study by Aspect Software Research determined that 44 percent of consumers would prefer to interact with a chatbot over a human customer service representative.
For businesses that can’t always respond to a website inquiry or phone call right away, the impact of implementing chatbots is huge: according to InsideSales research, an answer delayed by just 10 minutes will reduce a business’ chances to get the lead by up to 400 percent.
While consumer bots are already ubiquitous, business bots are just starting to take off. This trend will accelerate when businesses see that, using a combined decision-tree and AI-driven approach, business bots can nearly mimic human interaction in ways that consumer bots like Alexa and Google can’t.
Jeanna Liu’s love for nature is rooted in her childhood. As a young girl, Liu…
The arrival of generative artificial intelligence (genAI) into the mainstream at the end of 2022…
Data analytics and machine learning models deliver the most powerful results when they have access…
I’ve been on the road for almost a year now. Chasing freedom, adventure, and purpose.…
As technological use increases, so may the cost of innovation due to the global movement…
Have you ever asked yourself why some people are amazing at picking gifts, while others…
View Comments