The Hype and Reality of AI

Pure artificial intelligence, often featured in science-fiction movies and doomsday scenarios, aims to build machines whose overall intellectual ability is indistinguishable from or even surpasses humans. These machines will learn, reason, problem-solve, perceive, and understand language. But, even with early excitement and five decades of investment, only slow progress has been made. Pure AI still represents a hard problem.

However, applied AI, otherwise known as pragmatic AI, has a different history. Applied AI aims to produce commercially viable smart systems. Today, applied AI pervasively powers consumer experiences, making them simpler, smarter, and more strategic in our lives. Amazon and Netflix recommend products based on our histories, Yahoo and Facebook tag photos, Waze and Google get us to our destinations more effectively, Lyft and Uber precisely communicate arrival times of our drivers. Consumers are starting to expect intelligent experiences—personalized, contextual to their immediate situation, and highly relevant—from all of the companies with which they do business.

Applied AI is not a single technology. Instead, it's comprised of discrete technologies that either individually or in combination, are advanced enough to add intelligence to applications to learn, predict adapt and potentially operate autonomously. This intelligence produces quantifiable business outcomes that can be exploited today.

AI Will Transform Customer Service

AI can assist agents in completing repetitive, predictable tasks or completely take these tasks over. AI can interact with customers autonomously and with value. Yet, instead of replacing humans entirely, AI will enhance agents' skills and allow them to focus their attention beyond routine tasks, such as collecting and reporting information. Agents instead will be able to handle customer interactions that require deeper insight and analysis. These interactions will often take longer to resolve and are opportunities to nurture profitable customer relationships, which are increasingly rare in a digital-first world.

AI for customer service will do the followingl:

  • Deliver differentiated customer experiences. Close to half of consumers already use intelligent assistants like Alexa, Siri, and Cortana to sustain automated conversations. Intelligent agents for customer service will power single-purpose chatbots, such as what KLM uses to communicate booking confirmations, to virtual agents that embed deep learning. AI will make these conversations natural and effective and will delight customers. They will anticipate needs based on discerned context, preferences, and prior queries, and will deliver advice, resolutions, alerts, and offers. And they will become smarter over time.
  • Make operations smarter. AI will up-level contact center operations. AI will streamline inquiry capture and resolution. AI will extract useful information from voice and digital conversations, images, and machine-to-machine communications to quickly surface trends and customer sentiment that could affect customer retention and loyalty. AI will schedule maintenance appointments, push fixes to connected devices, and make field operations more efficient, for example, by restocking parts based on needs, or intelligently optimizing field resources to provide on-demand service.
  • Uncover new revenue streams and reinvent business models. AI can uncover patterns in large data sets that could reveal new insights that companies can use to create entirely new services for customers that can be monetized. Machine-learning algorithms used for business and customer intelligence find answers to questions that humans didn't even know to ask.

The following chart provides a few examples of pragmatic AI building blocks and what they can do for contact centers:

AI building block


Speech recognition

Converts the audio of spoken words to text that applications can use to take commands from humans (like Apple's Siri, Google Now, or Amazon Echo), transcribe a conversation, or participate in a conversation. 

Text analytics and natural language processing

Interprets typed text and whole documents. They classify words and parts of speech, apply linguistic rules and analyze relationships across an entire corpus. Advanced versions are used to understand emotions, sentiment, and, to some degree, intent.

Natural language generation

Is the inverse of NLP. It strives to express information stored and modeled in software in natural language that humans can understand.

Machine learning

Encompasses a set of analytical techniques that use algorithms to detect patterns in data, build segments, and create predictions. Machine learning can be used to either analyze data offline or to continuously build models to supply real-time context.

Robotic process automation

Performs routine business processes and make simple decisions by mimicking the way agents interact with applications though a user interface. RPA can be used to automate entire end-to-end processes, with humans typically managing only exceptions.

Image analysis

Identifies, and assigns textual labels to digital images and videos for identification and classification.


Kate Leggett is a vice president and principal analyst at Forrester Research, serving application development and delivery professionals.

Related Articles

Uber is testing a new customer support feature, but it's not for everyone. As first reported by TechCrunch, Uber is rolling out priority, 24/7 support capability for what seems to be a group of its top customers. The support programs include a hotline as well as an in-app feature, and specially trained experts will staff both channels at all hours of the day.

Posted November 17, 2017

Call intelligence company Invoca has introduced Signal AI, an artificial intelligence solution that leverages machine learning to gather insights from call center conversations with customers and feed them back to different organizational teams.

Posted August 08, 2017