As the “next big thing” emerges, it isn’t unusual for buzz to outpace understanding at first. Artificial intelligence continues to cement itself in the technological lexicon, but defining what exactly it is and how it can be used can be hard to pin down. Answers vary depending on who you ask.
 |
“It’s certainly a messy term,” says Stefan Weitz, chief product and strategy officer for Radial Inc. “I think of it as a measurement of how a system can achieve its goals in a dynamic setting. An AI system should be able to use its ‘intelligence’ to achieve a specific end game despite not being programmed for every obstacle it may encounter. Then it uses that knowledge gained to adapt for future interactions.”
“You’re trying to help people understand something they can’t understand by themselves,” says Jonathan Cherki, CEO of technology vendor ContentSquare. “AI is not machine-centered, it is an element of the machine, and it is using information gathered from human behavior,” he says.
“What we’re trying to do is automate what humans do when they interact and have AI take on human-level functions,” says Phil Gray, vice president of business development at Interactions, which uses AI for voice recognition and verification technology.
Related technology processes like machine learning add another layer to AI conversations as there are multiple types of technology and practices interwoven into the development and success of AI. Machine learning describes the ability to program an algorithm that teaches a machine what it should be looking for. “Imagine machine learning as a black box,” Weitz explains. “On one side you have all this raw data and on the other you tell the box what that data should look like. Then you let the box decide the best way to convert that raw data into analyzed data.”
Machine learning processes help refine artificial intelligence. When the AI-powered, virtual customer service agents from Interactions are presented with
a question or term it doesn’t understand, it transfers the inquiry to a human representative who answers the question [See First Person sidebar below]. The system feeds the question and answer data collected from the human-to-human interaction into its machine learning algorithm so the virtual agent learns and will know the answer the next time someone asks it. “Machine learning is the technology, artificial intelligence is the application,” says David Thomson, Interactions vice president of speech research.
In the broadest sense, AI can enable retailers to use their human employees for more complicated work tasks because they can entrust large and tedious jobs to AI processes. Radial, for instance, uses AI in its fraud-prevention tools. “For every order that comes through one of our [retailer] clients we are looking at over 1,000 variables to see if it shows any traits of a fraudulent purchase,” Weitz says. “That’s just too much for a human to look at.”
ContentSquare’s AI technology aims to help employees prioritize their operation and sifts through massive amounts of data to find specific issues buried within. For instance, apparel retailer Lacoste was having a problem converting sales on its newly launched mobile site. It used ContentSquare’s platform to analyze all browsing data and its AI-powered program identified specific pain points for shoppers pulled from the raw data. The analysis showed that half of users went directly to a subcategory page, such as visiting “Men’s shirts” rather than going through the “Men’s” page.
The analysis showed that consumers who visited a subcategory page directly were more likely to convert. ContentSquare suggested Lacoste highlight subcategory pages on the main menu.
AI capabilities can be applied across media. For instance, there is visual-based AI that uses technology to recognize images, which e-retailer Madison Reed incorporates into its Facebook Messenger chatbot (see page 27). AI can also be applied to written or voice-based interactions, a field called natural-language processing. The goal is to process human communication and create technology that recognizes common human speech or text patterns.
Interactions uses this type of AI to create automated customer service agents with human-like intelligence. Its recently launched Voice Biometrics program, for example, can authenticate a consumer’s voice so he can access his account, rather than answer a set of verification questions asked by a human representative. The company says it is targeting banks and payment services with the service.
Google, Amazon.com Inc., Apple Inc., Facebook Inc., Twitter Inc., Uber Technologies Inc. and other technology leaders have invested in applying AI in areas ranging from cloud services to customer service. Google began its Google Brain program in 2010 to further develop artificial intelligence using deep-learning techniques. Deep learning is another process of AI that tries to improve learning by mimicking human brain function. Google acquired DeepMind Technologies in January 2014 and has used the London-based company to expand its AI usage for Google Play, Ads and Shopping. The cost of the acquisition was not disclosed but analysts have estimated it to be between 400-500 million euros, or $425-$525 million.
In September, Amazon, IBM Corp. and Microsoft Corp. met with federal regulators to discuss the Partnership on Artificial Intelligence to Benefit People and Society initiative, which aims to put standards in place about how to properly use AI. The White House Executive Report that resulted called, “Preparing for the Future of Artificial Intelligence,” laid out the responsibilities of those in charge of developing and monitoring the technology as it continues to expand. The report says the government, “should monitor the safety and fairness of applications as they develop, and adapt regulatory frameworks to encourage innovation while protecting the public.” It also recommended that governments fund research on technology that could benefit the public good. Developers should take steps to “ensure that AI-enabled systems are governable; that they are open, transparent, and understandable; that they can work effectively with people; and that their operation will remain consistent with human values and aspirations,” according to the report.
The titans of technology are looking toward a future where artificial intelligence is a part of everyday life. For now, the amount of money and brain power being put into the application of artificial intelligence should put to rest any fears that it is a passing fad. The speculative buzz that once surrounded it has been replaced by the growing roar of the next great technology.
First person: Talking with an AI agent
I called the Hyatt Hotels Corp. reservations line to see how their artificial intelligence-powered agent would handle a relatively straightforward phone call. The system uses technology developed using Interactions LLC’s natural-language platform.
After a standard introduction, the voice asked whether I was calling to make a new reservation or adjust an existing one. “Make a new one,” I said. The next step was to locate the hotel at which I wanted to make a reservation. After I responded “Chicago,” it provided a selection of hotels. The process was pretty smooth, but it also asked me questions that it would have direct answers to.
The next prompt asked for my check-in date. I asked which day had the most vacancies, and this is when things started to go a bit haywire. The voice attempted to confirm that I wanted to check in July 2, 2017. I told the voice that was incorrect and was immediately directed to a human representative, who handled the rest of the call.
I wanted to keep testing the capability so I called back several more times, which produced similar results.
My conclusion is AI-powered agents are helpful as long as you stay on track with the prompts. Once you veer from the path by asking a question it isn’t prepared for, the system will quickly transfer the caller to a human agent.
Voice-based artificial intelligence customer agents are clearly still a work in progress. But for standard requests, AI-agents can help consumers avoid long wait times as they wait for the next available agent, and allow companies to reserve their human agents for more complex requests.