Tis the season for Listicles – those wonderfully quick reads with the “Top 10” things to look for in the coming year. Entering 2016, as Amy Stapleton noted in this post, the list of trends we were tracking was distilled down to three:

• Customers started showing a pronounced preference for self-service over lengthy interactions with live assistants.
• Conversational commerce, involving chatbots through mobile devices was rapidly expanding
• All flavors of Intelligent Assistants (especially bots and virtual agents) were making the transition from “assistant” to “advisor”

Indeed, these three factors shaped IA in 2016 and will continue to do so in 2017, spurred on by increased spending by business enterprises on Speech Processing, Natural Language Processing, Artificial Intelligence, Machine Learning, Knowledge Management and Predictive Interaction Analytics. It’s great news for the 28 or so firms that offer platforms for Enterprise Intelligent Assistants [link to our Landscape] while, at the same time it is a tremendous shot in the arm for the thousands of developers who are building bots for messaging platforms (especially Facebook Messenger) or “skills” for “metabots” (like Amazon Alexa or Google Assistant).

As these trends gain momentum and follow their own, respective, arcs I’d like to make this a “Wish Listicle,” exhorting the community of solutions providers to address some of the tough challenges that the rapid advancement of conversational commerce and intelligent assistance has exposed. Specifically:

Wish #1: A Nielsen-like rating service for the myriad of Chatbots

During 2016, the number of Chatbots on Messenger ballooned from zero to over 38,000 (especially impressive since the tools for creating bots for Messenger weren’t available until around May). We expect every platform for messaging and collaboration to witness similar bot onslaughts. Our own research shows that something on the order of 800 businesses or brands have implemented something north of 1,000 Enterprise Intelligent Assistants Virtual Agent Chat on the Web site or human-like assistants over the phone.

As it is with the plethora of apps in mobile marketplaces, a system needs to evolve to promote discovery and promotion of new bots, rating systems through which users can provide feedback and third-party analytics providers to aggregate statistics on real-world usage and popularity. My wish is that all of these come to fruition in 2017.

Wish #2: Speaker Identification for Families or Households

During 2016 the number of “skills” for Amazon’s Alexa expanded from 100 to 5,000 as the number of Echo-branded devices grew to an estimated 5.5 million. Alexa is proving to be the exemplar Metabot, a voice based assistant that can answer questions, play music, control household appliances and other tasks in response to a wake-up phrase and spoken instructions. It is proving to be quite popular and efficient, but it is not personal, secure or individualized because, for the most part, Alexa doesn’t know who’s speaking.

At any point in time, the speaker could be the primary Amazon account holder; or it could be a family member; or it could be a visitor with no connection to the account at all. This is a real barrier when you start to conceive of applications or services that highly personalized, private or subject to regulations that call for a modicum of security and protection.

Voice biometrics seems like a natural solution to the challenges of speaker identification and authentication. We recognize that identification of a single speaker out of a group, based on a voiceprint, can be a big challenge. It is orders of magnitude more difficult than matching a stored voiceprint with the spoken password of a known individual. But it should not be that difficult if there is a small group (3-6 people) of pre-registered individuals to choose from. Alexa could identify the speaker with some level of certainty. In many instances it could then tailor responses or prompts based on confidence that it is dealing with a specific individual. Then, if required in the normal course of an interaction, it may ask the individual to confirm his or her identity through the Amazon App on a smartphone or tablet. But it starts with satisfying my wish for “Speaker ID for Households”.

One of Amazon’s major contributions to Intelligent Assistance is the introduction of a device that can perform highly accurate speech recognition in a kitchen or family room. It can detect voices and distinguish between speakers based on where they are in a room. But

Wish #3: Help Cross the “Online/Offline” Divide

To date, interactions with Intelligent Assistants have been transactional in nature and purpose driven. Individuals turn to bots to speed up the process of carrying out a search, recommending an activity, booking a table or movie ticket or ordering a pizza. Though some people call this “Conversational Commerce,” there’s not much of a conversation going on. Likewise with Enterprise Intelligent Assistants which provide answers to specific questions or route people to resources in response to a query or “trouble ticket.” For the most part these interactions are “one-and-done” and success is measured by task completion, capture rates and call diversion.

IAs have an opportunity to transcend customary barriers between online and offline assistance. They can be omnipresent and achieve unprecedented levels of context-awareness that can be used for the benefit of each individual. In the healthcare domain, a “Wellness Coach” is a perfect example. Knowing an individuals medical conditions, indicated objectives for weight and activity, prescribed medications and regimens and other important data and metadata, a bot or IA defines a new ever-present and always-on consumption model for the community of doctors, hospitals, insurance companies, pharmacies, personal trainers and other care givers. It will be one that is under the control of the individual, in that it recognizes the his or her intent (based on NLU), but it is able to deliver alerts or spontaneously make recommendations based on applying deep neural networking, machine learning and analytics to the data and metadata made available to the IA.

Banks and financial services companies are following a similar model. Bank of America’s Erica is the most recent case-in-point. USAA’s Saving’s Coach is another example of efforts to make an Intelligent Assistant into a resource with understanding of one’s personal goals, empathy in one’s current situation, access to immediate feedback and ability to make suggestions like, “if you want to save for college, you might want to cut down on your spending for that second Frappucino at Starbucks.”

Many of us live life online and offline simultaneously. Think of all the people walking down Main Street with heads down and eyes glued to the screens on their smartphones. Although we think of IA’s as being “digital,” they have the unique opportunity to assist us in both domains.

Categories: Conversational Intelligence, Intelligent Assistants