How easily we interact with computers strongly informs how likely technology is to disrupt a given aspect of life or business. When we needed to punch code into a command line just to load a program, computers were far less user-friendly. But the mouse and graphical interfaces made things much easier, and computers blossomed from niche products into the mainstream. Touch took things further still, helping create a world where most people carry a computer in their pocket while increasingly also wearing one on their wrist.
What’s the next frontier that will further evolve human-computer relationships? Conversational AI.
You might be thinking that voice interfaces are nothing new—after all, smartphone assistants that you can talk to have been around more than a decade. But you’ve probably noticed those assistants have become better listeners, better conversationalists, and overall much more useful—and that’s because a range of technological breakthroughs have occurred behind the scenes, not only improving smartphone experiences but also inserting AI-powered voice technologies into a range of new devices and use cases.
For example, Google AI researchers opened-sourced BERT, a technique for natural language processing that makes voice models more context-aware and easier and faster to train. DeepMind, one of Google’s Alphabet siblings, also released WaveNet, which has helped create significantly more natural-sounding synthetic voices by replacing models based on phonetics with ones that use waveforms to predict which sounds likely follow one another. Both technologies are now deeply embedded in Google Cloud services such as Text-to-Speech, and they’re just a few among many examples of advances that help computers not only interact with us more naturally, but also act on our requests more effectively.
These recent developments mean our interactions with computers increasingly resemble our interactions with humans. Conversational AI can not only understand and naturally respond to our statements but can also be connected to other AI technologies, such as search or vision, to handle tasks we may otherwise only delegate to a trusted, qualified person. Soon, most human-computer interactions may not involve completing a series of set actions — clicking or swiping our way along a well-defined user journey — so much as just talking to machines and expecting them to keep up, even as the conversation changes course or topic.
Conversational AI is becoming a force across a range of technology categories and use cases, acting as a concierge who can speed up or automate aspects of our personal and professional lives.
Driving this trend are two dimensions to the way people—your customers—want to communicate with a business or public service: communication modes, and communication goals.
The concept of communication goals can make clear that speech acts, and thus the ways we verbally ask things of machines, are not all the same:
Information kiosk-type queries are generally the simplest interaction: one question (like “what time do you close?”) and one simple answer.
Information-seeking queries are more complex interactions, that can rely on a combination of speech understanding and traditional search engines operating over potentially billions of sources of information. A driver might ask their car navigation system, “What’s there to do in Breezewood” or someone cooking might ask their recipe app, “Can I substitute soy for salt?”
Requests for help are still more complicated interactions—things like asking to change the payment on a flight to frequent flier models or why your bill charged for 20 gigabytes even though you think you used only two. In these instances, several rounds of back-and-forth may be necessary. A short, respectful conversation with a problem-solving AI agent can be less stressful than waiting on hold or repeating information as you’re transferred from a person in one department to a person in another. In my view, whether your business can do these requests well will be a significant predictor of success over the next decade.
Full concierge interactions are generally the most complex interactions, with the AI going deep to solve complicated problems like “Why is my bicycle crank clicking?” or “What kind of cat should I get?” In these scenarios, the AI doesn’t start with access to the solution—it uses all kinds of business objects (i.e., over an Enterprise Knowledge Graph) to infer bespoke, even out-of-the-box solutions to queries. We are a long way down this path in many domains, and although I am unsure how fast all the supporting technology will progress, I am confident that excellence in “requests for help” queries is a major stepping stone.
Turning to communication modes, it is important to envision all the interaction models Conversational AI can encompass, and to be prepared to operate across all of them:
Texting
Talking on the phone
Talking within an app
Talking with an appliance
Virtual video conferences
Whether the use case involves interacting exclusively with an AI via an app, handoffs between an AI agent and a human agent over the phone, or transcribing video chats to extract actionable insights, the potential for improved service is significant—a fact we can see in the numerous forecasts for rapid growth in the Conversational AI market.
How your business can join the conversation
To meet customers where they are, your company must be available 24/7 on every channel. You’ll need multiple interactive modalities operating concurrently to tie it all together: e.g., talking to the customer, seeing what the customer is looking at, letting the customer view options, and make their selections, all in a conversation.
Building the AI backbone for these kinds of interactions can be expensive, difficult, and time-consuming. One way to help accelerate your progress is using open-sourced solutions, such as the aforementioned BERT, or vendor building blocks, such as Google Cloud products like Dialogflow, our Speech-to-Text API, or our Natural Language AI API.
Whatever your IT stack looks like, it needs to be both sophisticated enough to meet the anytime-anywhere expectations of today’s customers, and agile enough to adapt to whatever disruptions may occur tomorrow. This solution means capabilities like:
Automation & operational efficiency, e.g., AI-powered intelligence for contact deflection, predictive routing, agent productivity, etc.
Cohesive experiences across multiple smart devices, e.g., channel blending that supports multimodal engagement across apps, digital touchpoints, and modalities (mobile applications, website, phone, SMS over chat, etc.), with preserved context between touchpoints.
Data unification, management, and analytics, e.g., unified customer data on a single CRM or system of record.
Security, scale, reliability, and call quality, e.g., services auto-scale as needed, can protect personally identifiable information and other sensitive data, and are globally available with acceptably low latency across different devices and regions.
Conversational AI is a technology that enables machines to interact with humans in natural language. It has the potential to revolutionize how we communicate with computers and machines by providing an intuitive, conversational interface. With its ability to process natural language, it can understand user intent and respond accordingly. Conversational AI can be used in various applications, such as customer service bots, virtual assistants, chatbots, and voice-activated devices. This technology is becoming increasingly popular as it provides a more intuitive and efficient way of interacting with machines.
This technology is rapidly becoming an integral part of the customer experience, enabling businesses to interact with their customers in a more natural and efficient way. It can be used to create interactive chatbots that provide answers to customer queries and automate routine tasks, or virtual assistants that can understand complex conversations and act on user requests. With its ability to understand natural language, conversational AI has the potential to revolutionize how businesses interact with their customers.