In 1961, the world’s first industrial robot clocked in at a General Motors plant in Trenton, New Jersey. The 4,000-pound mechanical arm, named Unimate, was designed to weld cars and lift big pieces of metal. The robot was a huge success—even landing a spot on the Tonight Show with Johnny Carson.
Almost sixty years later, the emergence of artificial intelligence (AI) has seen machines leap from physical to mental labor. As computers step into roles that involve reasoning, a new wave of industries ranging from medicine to finance stands to benefit from—or get left behind by—AI.
Still, it’s not always obvious just what working with machines on a mental level looks like.
“Our ideas about partnering with machines on mental labor are much less specific because mental labor has historically been an exchange between people,” says Adam Pah, a clinical professor at the Kellogg School and associate director of the Northwestern Institute on Complex Systems. “A lot of our knowledge about robots is physical, and now that knowledge needs to be expanded to cover the realm of AI.”
Pah, who has collaborated with a number of organizations as they integrate AI into their operations, highlights three of the most significant ways that today’s AI can change how companies do business.
Learn What People Are Saying about Topics That Matter
People communicate mostly through natural language—not equations or Excel spreadsheets or structured daily reports. This has historically posed a challenge for businesses, because while the words we use are full of consumer insights, they are also generally inscrutable to analytical tools.
The ability to aggregate and quantify human language is, of course, changing. With natural language processing and machine-learning software, companies can now scrape vast libraries of text to generate insights on almost any subject.
It has now become commonplace for companies to analyze, say, customers’ Facebook comments to design more effective digital-marketing strategies. But organizations should be more creative and ambitious in thinking about how to apply insights from unstructured data, according to Pah.
Consider a cell-phone manufacturer launching a new phone.
Because these manufacturers make most of their sales through service providers such as ATT or T-Mobile, they are not directly connected to their customers. This means that post-launch customer feedback has historically been pulled from two sources—neither particularly ideal. The first is a slow trickle of information based on what, if anything, customers say when they return a phone to the provider. The second involves the manual monitoring of social media and tech websites by a small group of employees.
In a project he recently undertook in partnership with a cell-phone manufacturer, Pah proposed an alternative.
“We wrote a program that crawled the Internet for mentions on Twitter and Facebook along with about 30 sites where people review and talk about cell phones,” he says. “We then built a machine-learning solution that would automatically identify when the company’s product was being talked about, which features of the phone were being talked about, whether they were being talked about as positive or negative attributes. Then the program would group like statements together to give an idea about the magnitude of each problem—and it did this continuously.”
This new software allowed the phone manufacturer to start responding to issues immediately upon launch instead of after the typical six-week lag. It also provided a wealth of consumer insights—which features drove delight or disappointment, how people used the phone—that could be translated into future research and development.
The value of this AI technology, Pah notes, extends well beyond consumer-facing companies. A B2B organization curious about global-market potential could just as readily use machine learning to comb through analysts’ reports and assess economic conditions in certain sectors across the world.
“You can apply this kind of thing anywhere you have people publishing comments and reports at a scale that is too large to have someone read through every day,” he says.
The next step in this development, and one that is already underway, is training AI on images and video.
“That’s a really interesting twist,” Pah says. There is already a company, for instance, that identifies when a logo appears in pictures. “You can now expand your focus beyond what people write and track Instagram as a way to understand when and where you are showing up, as well as whom you should be targeting.”
Make Recommendations That Matter
Making recommendations is one of AI’s “oldest friends,” Pah says. Consider Amazon and Netflix matching customers’ past behavior with other products or movies they might like.
But despite the ubiquity of these recommendations, their reach is limited in an important way.
“So far, we haven’t moved much beyond this platform-centric world: some website recommends something else on the same website,” he says.
“Typically, there is a trade-off between the usefulness of an AI tool and the amount of privacy a user enjoys.”
This, too, is changing.
At the same time that machines are getting better at reading and interpreting data from a broad range of sources, personalized data is becoming abundant. Credit cards track customer purchases, web browsers log searches, social-media companies monitor likes and dislikes, and, increasingly, smartwatches and fitness bands track movement.
This convergence of analytical capability and data availability is fertile territory for AI to make more nuanced, specific, and wide-ranging recommendations.
Pah, for instance, recently received an email from TurboTax asking whether he would be interested in a machine-learning product that reviews tax returns, credit reports, and bank-account information to provide customized financial planning advice.
“That’s a step in the right direction,” he says. “We’re starting to recognize that we need to bring in data from everywhere to make more specific and useful recommendations.”
The next step in this development is a product that learns over time, honing its recommendations based on past behavior. Considering the TurboTax example, Pah suggests that this technology could track what financial advice people take and don’t take, and with what likelihood. With this knowledge, the company could then offer the advice that is most likely to be accepted first— potentially leading to greater trust in the “less likely” advice.
“Then it can look not only at what should be done, but what people are most likely to do, and start changing what it recommends based on this progression,” he says. AI applied in this way will not simply provide recommendations, but has the potential to do so in a sequence that influences human behavior.
“The key to that is bringing in a more complete picture of a person’s life,” Pah says. That more complete picture, however, is complicated by privacy issues.
“Typically, there is a trade-off between the usefulness of an AI tool and the amount of privacy a user enjoys,” he says. “AI systems need more data about a person to be useful, but information security is largely lacking at many major corporations and we have no strong protections or recourse as consumers when our data is stolen or lost.”
Provide Better Diagnostic Tools
In addition to finding relationships among key variables, AI can flag when certain complex conditions are met. These qualities make AI a particularly valuable diagnostic partner.
For an example in the medical realm, take the disease Acute Respiratory Distress Syndrome. Though its diagnosis is straightforward—the presence of four health indicators—80 percent of cases are not diagnosed because doctors and nurses in intensive care units are generally too focused on treating the primary issue from a patient’s admission to scan for multiple measures and connect the dots.
“Artificial-intelligence partners can really help in this setting because the syndrome has a basic definition and there’s no intuition to the diagnosis,” Pah says. “Where people may stumble, AI does this task consistently every time.”
As another example, Pah described the difficulty most hospitals have in measuring how effectively doctors use electronic medical records. Current metrics—how often doctors use the system, whether patients have access—tend to be rudimentary. But more sophisticated ways of measuring—such as whether the best treatment path was pursued—are difficult to generalize from one patient to the next.
“Accounting for all the small, unique parts of each patient’s life would take forever to build up using traditional models for everybody,” he says. “But it’s feasible to build up an AI system that has the ability to cover these kinds of things.”
And though the high-profile deployment of IBM’s Watson in healthcare has already drawn attention, the general premise is broadly applicable across many industries. Whether an organization is assessing new markets, deciding when to follow up with clients, predicting the retention rate of new hires, or identifying when mechanical parts are likely to fail, AI can signal the presence of certain conditions that people might overlook.
For instance, this diagnostic power can even be trained on the dreaded performance review. Quantitative assessments of one’s work can seem myopic and incomplete, so reviews are often frustrating experiences for employees. But where traditional analytics cannot account for all of the contextual variables that contribute to performance, AI can, provided the company can feed it the data it needs to make effective connections and recommendations.
“Where AI starts to excel is when you get a really wide range of variables to measure—when you have 30 or 40 columns in a spreadsheet that all need to be accounted for,” he says. “AI looks for patterns and relationships in all this information—projects, collaborators, success milestones—and relates them to what you’re trying to understand—an employee’s performance.”