How customers use financial services and engage banks is changing. The global pandemic increased frustration for banking customers as they tried to navigate legacy digital applications lacking personalization and waited in long queues to reach a customer service agent. According to the Capgemini World Retail Banking Report 2021, “Over the last 10 years, neo and challenger banks have attracted more than 39 million customers. The report finds that currently 81 percent of consumers said easy access and flexible banking will motivate them to switch to a new-age financial provider, in lieu of their traditional bank.”
Financial organizations need a solution that allows them to provide intelligent interactions with customers, as well as a friendly way to interact with customers to attract, retain, and provide them with a better experience. Natural Language Processing (NLP) uses Machine Learning (ML) which is a form of artificial intelligence (AI).
What is Natural Language Processing (NLP)?
NLP is a form of ML that enables computers to process text and speech, interpret the meaning of words, sentences, and paragraphs in context, and respond to human interactions.
How NLP helps in financial services
Machine Learning and NLP can help banks improve personalization and flexibility. Chatbots can handle many routine transactions with customers, speeding up their transactions. Chatbots can quickly understand if a chat session or voice caller wants information such as a balance update or payment confirmation. NLP can be added to live agent transactions to gauge sentiment and urgency of a caller. If someone is placing a call to report fraud, NLP can determine urgency and quickly move that caller to the anti-fraud department. NLP for language translation can be used when a language barrier exists between the caller and the agent.
Challenges faced by financial institutions and customers
A recent Forrester study shows that 84 percent of technical leaders feel they need to implement AI into apps to maintain a competitive advantage. Over 70 percent agree that the technology has graduated out of its experimental phase and now provides meaningful business value. These solutions include developing virtual assistants, chatbots, or messaging applications that provide customers with information or let them receive assistance from a virtual agent.
Many financial services organizations do not have the infrastructure that can handle processing and storage of the massive amounts of data generated by ML and NLP applications. To make AI and ML a core component of their business, organizations need faster, responsible ways to implement AI and ML into their systems–ideally using the existing skills of their technical team.
Financial service developers often do not have the technical skills needed to create ML models that incorporate NLP to create custom-designed digital voice assistants (chatbots) that can talk with existing or prospective customers. In fact, 81 percent of technical leaders surveyed in the Forrester study say they would use more AI if it were easier to develop and deploy.
Take advantage of cloud and GPU NLP-enhanced AI services
Microsoft and NVIDIA have a long history of collaborating on hardware and AI solutions that support ML and NLP features. The companies are working together to enable a close integration between Microsoft Azure and NVIDIA AI solutions to advance AI for financial services. Microsoft Azure AI Services such as Chatbots, Neural Voice and Language Understanding are built to enhance development capabilities through API calls to those Azure services. These ML services are enabled via NVIDIA GPUs and HPC compute resources via Azure.
Using NVIDIA’s AI platform—including GPUs speeds up ML, AI model training and inference—which empowers institutions to improve data-backed decisions and security, and enhance customer experiences. At the heart of conversational ML are ML models that require significant computing power to train chatbots to communicate. Once models are trained, the bots need to be able to engage in life-like conversations with customers in real-time meaning the models in production must operate at low latencies.
NVIDIA’s AI platform for finance speeds up ML model training and inference. On large NLP models, inference runs 10X faster on GPUs than CPUs. The performance and compute power needed to train the ML models are enabled by NVIDIA Tensor Core GPUs and supporting SDKs and libraries systems to efficiently use GPUs.
As part of the collaboration, Azure pre-installs the NVIDIA libraries to save time and operating costs. Microsoft and NVIDIA also offer Azure confidential computing with NVIDIA GPUs for trustworthy AI for sensitive financial system data.
Tools that make NLP and speech possible
Microsoft Azure AI includes a portfolio of AI and ML services that provide access to high-quality vision, speech, language, and decision-making models. Using Azure AI and ML tools lets developers use simple API calls to create organization-specific machine learning models with tools like Jupyter Notebooks, Visual Studio Code, and open-source frameworks like TensorFlow and PyTorch.
Developers can use Microsoft Azure Cognitive Services technology that includes a neural text-to-speech capability, which can be deployed anywhere from the cloud to the edge with containers. Cognitive Services brings AI within reach of developers and data scientists by providing built-in AI models and use cases. All it takes is an API call to embed the ability to see, hear, speak, search, understand, and accelerate advanced decision-making into an application.
Other Microsoft solutions that enable NLP in applications include:
- Microsoft Bot Framework used to create a chatbot with the ability to speak, listen, understand, and learn from users using Azure Cognitive Services.
- Microsoft Custom Neural Voice neural text-to-speech technology can customize a voice agent for an organization. Organizations simply record audio samples and upload training data. Microsoft will create a unique voice tuned for the voice agent recording.
- Microsoft Language Understanding creates multilingual, customizable domain-specific keywords, or phrases across 96 languages. An AI speech model can be trained in one natural language and this model can be used in multiple languages without retraining.
NLP-enabled AI optimizes financial service customer interactions
Customers now demand flexible online services to quickly meet their financial needs. Microsoft and NVIDIA compute resources provide NLP-enabled ML to enhance customer service and provide support through speech recognition, translation and understanding voice. The Microsoft-NVIDIA solution helps organizations improve performance, increase their ROI, and better serve customers.