Artificial intelligence that’s indistinguishable from human intelligence hasn’t arrived yet. But early examples of AI are sprouting up in a variety of industries, with businesses using it to improve operations, cut costs, and connect with customers. As more businesses turn to the technology, the market for AI products and services will grow to $36.8 billion in 2025 from $643.7 million this year, research firm Tractica projects.
Here are eight examples of AI at work today.
1. AI bank tellers
Banks are using AI-assisted programs and apps to improve customers’ online experience, including more intuitive interfaces, more effective customer service, and more secure transactions, according to an August report by Jost Hopperman, vice president of Banking Applications and Architecture at research firm Forrester.
For example, DBS Bank claims its Digibank app understands spoken customer requests to check balances, find interest rates, and make payments, among other tasks.
2. Self-driving car
A self-driving car requires a lot of sensors to accurately perceive and respond to its environment in real time. Automakers process data collected from the sensors to build and adjust the predictive models guiding these real-time decisions, such as acting appropriately in a construction zone, or braking to avoid an obstacle. BMW, Google, Toyota, and Tesla are among those pouring millions of dollars into efforts to develop autonomous vehicles, many of which are already on the road.
Because a car can’t rely on communication with a cloud-based computing system for an instant answer, it needs to be able to make decisions itself. Its development incorporates a common architectural aspect of AI: model training in the data center, followed by deployment on edge devices.
3. Air travel
Airlines are increasingly using deep learning neural networks and machine learning systems to optimize routes and resource use. With multiple destinations plus complications like traffic, weather, and fuel costs, defining or altering route plans is a complex task. London-based EasyJet reportedly uses AI systems not only to plan its flight schedules, but to manage supplies onboard for improved customer satisfaction and revenue.
Improving AI technology is helping these organizations respond to unexpected events in real time.
4. Lowering hospital readmission rates
The University of Pennsylvania Health System uses analytics and machine learning in its Penn Signals program to identify inpatients who, upon discharge, are at risk of readmission due to heart failure. The system, which Intel Chief Data Scientist Bob Rogers characterizes as early AI, correlates a diverse range of medications and health factors and sends a text notifying doctors to take additional steps for patients it flags, even if no formal diagnosis has been given.
Health system officials say the system has identified about 20 percent more patients who need added services before being discharged, improving care for the patients and saving money for the health system. (UPHS has partnered with Intel to help create and refine this system.)
5. Fraud detection
Credit card issuers, telecom companies and others use neural networks to analyze transactional and behavioral data, looking for patterns that indicate fraud. According to Visa’s Viewpoints blog, the credit card issuer processes more than 82 billion transactions per year, and aims to analyze 100% of its data to improve its models for making real-time fraud decisions, looking at hundreds of criteria including geo-location and purchase history. Fraud techniques change over time, as criminals look for weakest defensive spots, which machine learning is uniquely capable of keeping up with.
6. Reading radiology images
Radiology images require great expertise to read correctly. A medical program applying machine-learning algorithms as it scans a large number of images can discover patterns or degrees of subtlety that humans miss, thus improving the accuracy of diagnoses and reducing the number of required follow-up tests.
It’s an early example of the kind of projects that could soon be commonplace in healthcare. In hospitals and medical-research labs, “AI is being used to tackle that data overload, and extract information in a way that’s actionable,” says Venkat Rajan, global director of the Visionary Healthcare Program at business consultancy Frost & Sullivan.
7. Buying, selling, and displaying ads
More than two-thirds of all U.S. ad buying on mobile channels this year is being run through “programmatic advertising,” according to research firm eConsultancy, using AI to analyze audience demographics and ad performance to help buyers distribute, adjust, and monitor spending.
Using platforms such as DataXu and MediaMath, buyers engage in real-time bidding for ad space. They also retarget or customize ads based on cookies indicating that an individual has visited a particular page or already seen a particular ad. Retargeted ads were than other ads in an online campaign by carmaker Mazda.
8. Chatbots and virtual assistants
Retailers, banks, and others are using virtual assistants as a conversational interface for text-based customer interactions—in other words, chat. For example, according to HSBC, its chatbot ‘Olivia’ can answer potential customers’ questions about credit card and account offerings. AI is providing rapid advances in natural language processing, though there’s still room for improvement (see Chatbots article.)