Blog Financial Institution
Artificial intelligence (AI) is poised to be the next disruptor in the financial services industry. In fact, experts predict AI will net FIs $1 trillion in savings over the next 10 to 15 years, reduce operating costs by 22% and spur a 34% increase in revenue growth. However, the world of AI can seem like venturing into uncharted territory. What does all that tech jargon mean? Get up to speed with this primer on need-to-know AI terminology.
A: Artificial intelligence is the term given to a number of techniques that enable computers to emulate human behaviors. Examples include interpretation, understanding, reasoning, planning, communication, machine learning, deep learning, natural language processing and intelligent automation.
The typical AI toolkit includes multiple “behind the scenes” elements to make it work: algorithms, neural networks, deep learning and supervised learning. Each is defined below.
A: Machine learning is the general term that describes the process of getting a computer to act without being specifically programmed. The computer instead learns by experience and without human intervention.
For example, fraud detection applications can scan millions of documents to identify red flags and alert banking staff. Over time, computers discover patterns that suggest fraudulent activity – effectively learning to become better at their jobs.
In another example, AI-driven banking platforms can monitor customer behavior and transactions. Over time, they can learn to match banking products to target customers and identify the right time to make a personalized offer. They can then trigger chatbots or virtual assistants to deliver upsells and cross-sells when customers are the most likely to act.
Machine learning is not the same thing as AI; rather, it’s one of AI’s many subsets.
A: Algorithms are the basic building blocks of artificial intelligence. They’re simple “recipes,” or rules, that make AI work. Essentially, algorithms are step-by-step instructions such as programming commands and math formulas that instruct computers on how to solve problems on their own using a specific set of inputs or “ingredients.”
For example, an algorithm can be written to scan remittance emails that arrive separately from payment data. They can then extract unstructured data such as vendor names and payment amounts from those emails and automatically match them with ACH payments. In this case, an algorithm can save staff from spending tedious hours on manual data entry, speed up straight-through processing and save the bank money. When you implement integrated receivables, you also have a compelling product that can be used to strengthen relationships with corporate customers and create a new revenue stream.
A: A neural network helps a computer develop human-like functions, such as perception, reasoning, visual recognition or language processing. It organizes dozens to millions of artificial neurons called “units” into layers, where a different type of processing occurs in each layer.
As data and inputs are passed through successive layers, the neural network develops greater understanding – just like humans use multiple sense and types of thinking to interpret the world around them. Thus, a neural network simulates the sophisticated hierarchies and connections between neurons in the human brain.
Neural networks are employed in a variety of AI applications for banking. Chatbots are one example. When a customer engages with a chatbot, one layer of the neural network might recognize the language being used. Another might decipher the meaning and context of a typed question. Yet another might access data from multiple sources to find an accurate answer. The next layer might translate that answer into a meaningful response in the customer’s native language. As a whole, the neural network can hold a human-like conversation that helps the customer get the help they need.
A: Deep learning is what occurs when data and inputs pass through the neural networks. It’s what separates AI from traditional computers. Where traditional computers must be programmed to follow specific instructions at each stage, deep learning is autonomous and self-teaching. Each time an application completes a task, it finds patterns and improves performance.
Deep learning plays a role in many different AI applications. Personalized offers are an example we previously touched on. For example, your banking platform might monitor customer activity and discover that small business customers often take out business lines of credit. The system might automatically deliver offers to apply for credit to small business owners.
It gets even better: the system might try a variety of chatbot greetings to introduce the offer to customers. Over time, it might discover that certain greetings have higher conversion rates. It might also discover that greetings can be optimized for different types of small business customers based on things such as industry, account balances and demographics. In this example, you can see how deep learning can be applied to not only help customers get the best services for their needs, but also maximize revenue for your bank.
A: Even though AI powers autonomous computing, applications must still be “trained” to do their jobs well. Supervised learning describes the most common approach to training an AI application. Using a training set of data, the organization provides the computer with both the question and the answer.
Self-driving cars offer a good example, since they must be “taught” to recognize traffic signals. Human trainers present self-driving cars with a question: “Is this a signal to stop?” The answers might include a set of “yes” images clearly labeled stop sign and red light, and a second set identified as “no” images. The application learns to recognize when the car is supposed to stop.
With unsupervised learning, organizations provide the question without the answer. Instead, they apply advanced programming that shows the computer how to carry out complex learning tasks on its own. It’s a less common approach, since it requires significantly more upfront work; however, systems capable of unsupervised learning represent the full potential of artificial intelligence.
A: NLP is a function of AI that trains computers to interpret and respond to human communication in text or speech forms. You’re probably already familiar with NLP, as the technology is used for popular virtual assistants such as Siri and Alexa.
For banks, NLP can power chatbots and virtual assistants, which can help customers get instant, accurate answers to questions and self-service their accounts. In addition to humanizing machine communications, NLP enhances the customer experience by granting convenient and personalized access to their accounts.
And, since virtual assistants and chatbots can handle routine and redundant customer service, NLP helps free staff time to focus on high-value, revenue-generating activities such as deeper conversations with customers.
A: Image recognition enables computers to identify objects, places, people and handwriting that exist as images. Sometimes referred to as computer vision, image recognition powers technology you’re already familiar with: smartphone facial recognition, for example, and self-driving cars that recognize pedestrians in the roadway.
Customer authentication is one way banks can employ AI-powered image recognition. It can be used to verify customer identities when they log into their mobile or desktop banking portals and even at bank branches. Imaging recognition can also be used to verify customer signatures on important documents.
A: Chatbots leverage natural language processing to conduct interactive “chat” conversations with human users. Also known as conversational interfaces, chatbots can be deployed on websites, mobile apps and telephone systems.
Banks commonly use chatbots for customer service. As previously mentioned, chatbots are often the user-facing end of a series of technologies that allow applications to deliver data-based, personalized offers, upsells and cross-sells customers are likely to act on.
For example, let’s say one of your customers often makes international wire transfers. Rather than try to upsell them with hit-or-miss banner ads, your portal could greet them with a chatbot that says:
“Hi Joe! I noticed you made five international transfers last week. Did you know there are other payment options available at a lower cost?”
Naturally, the customer is more likely to respond to the chatbot prompt than click a banner ad. The chatbot can then make an immediate upsell or pass the warm lead on to a banker to have a deeper conversation.
A: Virtual assistants are AI-powered applications that help customers complete common tasks. They power self-service functions. For example, customers can use virtual assistants to check account balances, transfer funds, find an old statement or locate the nearest ATM.
Virtual assistants can be combined with other technologies. For example, they can be deployed via chatbots that allow customers to self-service their accounts via text and voice commands. They can also connect to the neural network to draw deep insights into customer behaviors and market trends and offer sound, data-backed financial advice. In this capacity, they’re known as robo-advisors.
When banks implement virtual assistants for customer service, they’re able to quickly provide customers the help they need and free staff members to focus on high-value activities that generate revenue.
A: Robotic process automation refers to specialized software that’s programmed to handle routine business processes. Though RPA isn’t technically a component of AI, it’s a common first step before AI initiatives are implemented into banks.
RPAs can be used to collect data, update spreadsheets and move information between applications. They can quickly and accurately handle repetitive tasks and basic workflows. Like many other technologies, RPAs lessen the burden of manual data collection and entry. They save money and free resources for more valuable endeavors.
A: Intelligent automation is a more advanced form of RPAs that’s used to streamline business processes. Also known as intelligent process automation, machines not only take over routine and repetitive tasks typically handled by humans, they also leverage AI to learn and, over time, do them better.
Integrated receivables are a prime example. Remember how AI algorithms can be used to automatically match “stranded” receivables to remittance data? With intelligent automation, such platforms can learn over time and significantly improve performance, so their value grows exponentially.
A: The top uses cases for AI in banking are:
Artificial intelligence can bolster fraud detection initiatives. Credit card processors that instantly scan millions of transactions are one common example, but AI can also enable banks to automate Know Your Customer (KYC) and Anti-Money Laundering (AML) compliance.
The same technologies can be used to make better credit and underwriting decisions by analyzing both traditional and non-traditional data like job history and education to help bankers determine whether a loan applicant is a good candidate.
Chatbots and virtual assistants make it easy for customers to self-service their accounts and free bank staff from redundant, time-consuming tasks so they can focus on stronger customer relationships. These enhance the customer experience through instant access, convenience and trust.
Biometrics and facial recognition also allow instant account access and eliminate the need for passwords and access codes, a source of frustration for some customers. Overall, AI eliminates friction points and enables banks to foster meaningful relationships and long-term customer loyalty.
AI applications automate tedious, time-consuming and costly tasks. They can handle integrated receivables that increase straight-through processing rates by up to 95%, contract reviews, reporting and workflow automation.
Ultimately, artificially intelligent solutions eliminate the need for tedious manual tasks, improve accuracy and compliance, and save banks money.
Chatbots and virtual assistants that deliver personalized offers, upsells and cross-sells have higher conversion rates than live agents. Integrated receivables create a compelling product that can be sold to corporate customers.
AI applications can also monitor accounts for red flags that suggest a customer is about to leave your bank. For example, if a customer withdraws large amounts of money and logs in less frequently, the system can alert a banker about the at-risk customer. The banker can then contact the customer, resolve issues, save the account and stem attrition.
AI is responsible for a seismic shift in how banking is done. The sooner you familiarize your bank with AI terminology, the sooner you can begin having conversations about how your bank can harness its power to reduce costs, enhance the customer experience and grow revenues.
Blog Financial Institution
Blog Financial Institution