Conversational AI

I have recently had the opportunity to be interviewed for a number of articles and podcasts for my perspective on Conversational AI.  A compilation of many of those questions and responses follows.

– Matt Smith

What is conversational AI and how is it unique from other forms Artificial Intelligence?

Conversational AI refers to a very human-like, personalized way for people to interact with technology.  It can accomodate a number of input “channels”, such as websites, mobile apps, text messaging, voice calls, intelligent speakers, even email.

A year ago, most discussions about Conversational AI were very technology centric, and an alphabet soup of abbreviations, (STT, TTS, ASR, NLP, NLU).  Today a lot of people just refer to all that technology as “the AI”, as in, “Someone can ask a question and ‘the AI’ will answer with the right information.”  

You’ve described “Conversational AI” as ‘hands-free computing’ or ‘screen-less computing.’  How is this changing the way brands compete for customers?

First, Conversational AI is a new user interface (UI) as much as a set of technologies.  It replaces keyboards, mouse clicks and screen taps with voice instructions, as when you tell your navigation app your destination rather than typing in an address.   That’s where the idea of hands-free computing comes from.  More and more, it’s also a screen-less experience; smart speakers, voice enabled appliances and many in-car functions are examples.

What this means for brands is they now must learn to compete more on customer experience (CX) rather than visual representation of brand or product.  Competing on CX means developing differentiation around characteristics such as ease of use, convenience, personalization and contextualization, all of which can be enabled or optimized with Conversational AI capabilities.

What are some of the best opportunities for Conversational AI you have seen?

Many companies start by targeting the “informational request” categories on their websites, customer portals or mobile apps.  Informational requests examples include FAQs pages, forms or documentation searches, store locations and hours, or product instructions, like ‘how to reboot your Wi-Fi router.’  Processes with the following characteristics are an ideal place to begin with Conversational AI:

  • High volume
  • Highly repetitive
  • Lots of historical data
  • Widely applicable
  • Easily measurable

After bringing Conversational AI to all those FAQs what is the next category to target?

Level two for Conversational AI are transactional requests.  These activities allow customers to actually complete an activity or task, such as opening an account, transferring funds, getting shipping status for an order, or changing a scheduled service date.

Solving transactional requests in Conversational AI are more complex, as they generally require deeper application integration and effective personally identifiable information (PII) management.  Also, customer expectations are higher as the interaction is now a highly personalized transaction.  In terms of creating greater customer value though, transactional requests are where Conversational AI’s highly personalized and contextual experiences become a significant differentiator for brands.

What are some Conversational AI examples your team has built?

We helped one of our airline clients replace a website FAQ page with a natural language chatbot to help manage loyalty program miles.  The teams started with the 50 top requests they received, compiled past response logs, analyzed the data and mapped the best answers to each question, then launched the new chatbot after a round of internal user testing.  Within a month of launch they were taking thousands of requests through the chatbot and resolving more than 80% of them without any escalations or agent interactions.

Here are a few more projects we’ve recently finished:

  • We built a drive-thru solution for quick serve restaurants to speed the ordering process and improve accuracy.
  • For banking clients we’ve created a virtual mortgage advisor that helps home buyers research loan options.
  • We developed digital assistants for automotive companies to help car shoppers compare vehicles and manage vehicle maintenance scheduling.

Has Conversational AI officially become a mainstream strategy for most brands?

I believe Conversational AI is entering its third phase.  It has progressed very rapidly over the last few years to where it is now, and adoption will only accelerate as more consumers embrace and expect the kind of personalized experience Conversational AI delivers.

What is the biggest factor driving adoption of Conversational AI among your clients?

I see a dual mandate driving Conversational AI’s adoption right now, which is for brands to better serve both the commerce and care needs of their customers.  Keep in mind, customers in this context can refer to actual “paying customers” or it can apply to visitors, leads, members, patients, users, etc. 

The commerce component of this dual mandate is about driving lead conversions, increasing order size or volume, improving customer retention, encouraging repeat buying and enhancing any other form of revenue generation.  The care side of the mandate applies to service, support, help, how-to’s, instructions, guidance and recommendations. 

Every industry has faced this dual mandate since the first days of human to human commerce.  What’s changing in this Conversational AI era is many of the old rules and best-practices must be unlearned as completely new strategies now exist for engaging, selling and supporting consumers in a hands-free and screen-less world.

What will be the biggest challenge for Conversational AI in the next year?

Bot Backlash is easily the biggest challenge brands will face this year, but it’s also opportunity for us all.  Bot Backlash will happen as consumers increasingly push back on chatbots and virtual agents that are just a poorly executed version of, “Press 1 for tech support, press 2 for sales.” What will prevent Bot Backlash is grounding Conversational AI in a customer first philosophy, and delivering the right design, technology and personalization to create an experience consumers prefer over searching through websites to navigate an FAQ or calling a contact center agent for help.

Will Conversational AI eventually extend beyond traditional computing devices?

CES 2019 showed us exactly how fast Conversational AI is spreading beyond traditional technology devices.   This year’s show featured smart homes, vehicles and appliances; workplace voice solutions for accessing buildings, meeting rooms, data and systems; virtual agents for physician offices, hospital rooms and ORs; and conversational AI helping customers at drive thrus, hotels and airports.  Given how pervasive it’s already become, it’s hard to believe that we’re still in the very earliest stages of Conversational AI.  Alexa isn’t even five years old yet (it was released in November 2014).  We can only imagine the things we’ll be doing in just a couple more years.

Finally, will Conversational AI really help employees or is it just another job killing technology?

Conversational AI is a great ‘work assistant’ for employees in many ways.  For example, it can work in the background with contact center agents to access caller history and increase accuracy of responses to help requests.  Retail sales agents can use Conversational AI to search inventory, product information or pricing specials so they can focus on providing shoppers more personal attention.  In healthcare it can reduce the time physicians and clinicians must spend creating and updating electronic medical records. 

Nearly everywhere that today we find employees tied to keyboards to do their jobs, Conversational AI will increasingly mean that work can be completed hands-free and closer to the consumer.  Soon, workers of all roles and levels will consider Conversational AI as indispensable as their smart phones and wonder how they ever got along without it.  That’s exactly the level of experience we are striving to build today into every engagement.