William Meisel
January 9, 2020
Originally published by Speech Technology Magazine
(More on this topic at the Conversational Interaction Conference Feb 11-12, organized by Bill in his role as Executive Director of the industry organization AVIOS.)
General digital assistants like Google Assistant and Apple’s Siri on smartphones and Amazon’s Alexa on in-home devices are competing to be portals to the web. These portals support individuals talking to companies that have developed company-specific “skills” for Amazon and “actions” for Google that can handle customer support and even convert an inquiry into a sale. Apple has a similar option, “shortcuts,” supported by SiriKit; the Apple option is more task-specific, e.g., placing a delivery order at a restaurant.
Companies must eventually be available through these portals. In my recently published Computer Intelligence: With Us or Against Us?, I argue that company-specific digital assistants will be as necessary as web sites in the relatively near future. I further argue that, once a customer contacts a company for customer service and is interacting conversationally with a company assistant, that connection can be an opportunity for advertising and sales. The company digital assistant can also provide additional information and entertainment, such as recipes using a food product. Because of this expanded content, even companies that have a minimal need for a customer service line will find they need a friendly digital assistant.
The portals drive the need for company digital assistants reachable through them, but companies can also make their digital assistants accessible outside the portals. Users can chat by voice or text through chatbots on web sites or download mobile apps, for example. This option expands customer service applications well beyond that single task.
Bank of America’s conversational mobile app Erica is an early example of a direct mobile app. In mid-2019, the bank announced Erica was helping more than 7 million clients and had completed more than 50 million client requests, adding an average of 500,000 new users per month. Erica allows interactions such as the following:
• Disputing a charge;
• Obtaining a snapshot of monthly spending;
• Finding balances, deposits, holds, purchases, refunds, and credit card transactions;
• Exploring home equity lines; and
• Getting info on Preferred Rewards.
The interaction with Erica is in natural language. It allows intuitively stated requests, such as the following:
• How much did I spend on Saturday?
• What’s my available credit?
• Did I have any grocery purchases over $100?
This extended example suggests that the underlying technology is mature and flexible.
Automated conversations using natural human language—speech or text—is a technology that has passed the tipping point of utility and will continue to get better at an accelerating pace. However, like web sites, company digital assistants can be poorly designed. Interaction through human language must tolerate many variations in the way a user might request information. Companies must develop digital assistants carefully using staged deployments to learn what customers will ask.
A careful effort to build effective digital assistants, as a result, has a significant side benefit. It forces companies to better understand their customers’ issues and needs. In addition, customer service that used to be perceived as a frustrating cost of operation can become a marketing channel.Interactive ads enabled by conversational technology do more than just repeatedly display the same message and will become the norm.
Building company digital assistants doesn’t require a major research project. The general digital assistant providers offer free toolkits for building skills, actions, and shortcuts that support templates and other aids that allow focusing on content. Further, hundreds of independent companies are in the business of supporting creation of digital assistants, both for the general assistant portals and for independent deployment.
Companies shouldn’t wait to begin experimenting with the technology until they begin losing business to competitors. Creating effective natural language deployments—like deploying effective web sites—takes time.
(More on this topic at the Conversational Interaction Conference Feb 11-12, organized by Bill in his role as Executive Director of the industry organization AVIOS.)
