The Way to Achieve the Most Out of Your Banking Money with AI and Machine Learning

AI is already implemented in three-quarters of the banking industry. And even though Gartner forecasted that GenAI will become one of the essential banking trends only after several years, as of late, there has been an apparent urge to implement GenAI in the mix right now. The contemporary technology is not about locating the most ideal tool. It is about the selection of a tool and putting it in the proper place, searching the appropriate business case and then utilizing and abandoning the technology. Technology is evolving at such a pace, it has already become a disposable aspect, and your next solution is probably not going to last you more than two years.

Particularly in light of the business implication. For instance, the primary purpose of AI in banking is to automate tasks and generate predictions using machine learning (ML) models (including complex deep learning). This needs huge computing capabilities and data, and substantial investment. Financial institutions are also among the heavy consumers of AI already. They rely on cloud-based ML services (e.g., AWS, Microsoft Azure, Google ML) and (mainly) private or mainframe hybrid cloud environments. It can appear that there is a new technological breakthrough and almost every day, and naturally, everyone wants to utilize it right now. Still, a lot to be gained by evaluating what you have to do in real terms and what solutions can assist you to attain that. And by making sure that you are working with what you already possess to the fullest extent possible, getting the most out of it, in other words, making the most out of your money, before you begin spending.

Already, three-quarters of financial institutions employ AI and the recent progress presents optimism regarding the growing interest toward introducing GenAI to the technological mix as early as possible. The low-code/no-code software developed by GenAI may save a lot of costs, such as the reduction of operational expenses and risk associated with the non-reliability of the old infrastructure and work-consuming processes.

The key AI tendencies in the banking sector, as reported by Gartner:

• Computer vision: Improve operational performance in KYC, AML, ID verification, fraud detection and underwriting
• Decision intelligence and graph technologies: Modernize front- to back-office applications
• Groundbreaking AI functionalities: Increase precision and extensiveness in risk management and marketing through foundation models. Generative adversarial networks (GANs) are well applied in the creation of realistic images, videos, voice recordings that are in most cases indistinguishable to the original information
• Model-centric AI: Introduce encouraging developments such as composite AI and GenAI to support fresh initiatives
• Data-centric AI: Focus on data analysis for ethics, accuracy and explainability
• AI-based applications and use cases: Develop conversational user interfaces, smart spaces, robots, etc.
• Responsible and human-centric AI: Double down on ethical AI development, risk management and beneficial social impact

Decision-making, operations management, and product hyper-personalization to increase client returns would bring in a lot of improvements to your operating revenues and cost reduction. S&P estimates that a 10 percent decrease in the cost of the bank staff would boost the returns on equity by approximately 100 basis points and cost-to-income ratios by approximately 3 percent (according to the S&P Global Ratings, Global Top 200 rated banks).

Artificial intelligence strategies are potentially capable of offering competitive advantages to banks who have the capability and flexibility to utilize it best. It merely requires some foresight and a wise implementation of AI. As the upscaling of risk management takes place, the AI might influence the perception of bank risks profiles incidental to it. When banks are able to price credit risk adequately by using patterns latent within the information to compute the probability of customer debt repayment, the banks will improve their workout models, minimize vexing loans as well as maximization of correctness. The management of risks could be enhanced through the use of AI in a number of ways. Nevertheless, it will not prevent poor implementation to create reputational and operational risks as well as a negative risk position to the bank. Get it right. Find a technology provider who understands the banking ropes.

The new AI Act in Europe has created an all-encompassing control of AI products and services in a risk- and use-case-based setting. The Act comprises the foundation model responsibilities in terms of general-purpose AI systems. Other countries such as Canada, Brazil, Chile and the Philippines are not left behind in such moves. In the United States, the National Institute of Standards and Technology (NIST) has established an AI Safety Consortium. It seeks to collaborate with industry to come up with new standards, such as industry-specific guidance, which is risk management-led. As well as governmental regulatory activity, there's an increasing acknowledgment of industry's duty to protect society by initiating or strengthening AI governance.

The novel AI regulation application is founded on the risk that is likely to be introduced by risk users and the activities that should be controlled. There are four types of risks that have been taken into consideration by EU:

The words minimal and limited risks mean that AI is employed in simple internal functions such as automatising routine. They do not involve communication with real users and customers (e.g., bots or bots related to the knowledge base). Minimal risk does not require much regulation, and the majority of the regulation that will be used is concerning letting customers know that they are dealing with bots. The risks that are deemed to be high and unacceptable include AI systems which inform decision-makers. This consists of query scoring, matters pertaining to client data and dealing with clients in totality. These AI will be highly controlled, and a lot of regulations will have to be taken into account using these kinds of AI systems. The use of AI systems and operations that voluntarily or unwillingly intervene with users by modifying free will or behaviors will be prohibited. In this way, the seniors and minors, the functioning of the society, such as social scoring and biometric screening, and the analysis of the data with the presumptive participation of the customers, including their origin, activity, etc., will be impacted. Plans that are implemented in 2024 and implemented by 2026 will be mostly applicable to high-risk systems.

Financial services

Access will be based on AI regulation in financial services. Financial services In finance, AI systems analyze client data to score their credit worthiness and detect fraud.

Analytics wise we are in a period of exponential development in this spectacular technological era. Thus, everything constructed just 2 years ago is already obsolete. And the issue is how can a company utilize the old technology in a manageable, non-toxic manner with consideration of the knock-on effects of the tiniest mistake? Take ChatGPT, for example. Nobody is aware of the source of information in the app (or its bias/copyright/integrity). Nonetheless, an increasing number of employees are getting incentivized to utilize the application to create corporate blogs, articles, white papers and even code lines. No wonder financial institutions are considering building ringfenced AI services (an extension of the ongoing company debate: "Do we go to the cloud or stay on-prem?"). It becomes a serious problem since, to have an effective AI and advanced analytics, you have to be confident that your data integrity is up to par. With that said, it is only logical to work in-house, offline. The company can only train your corporate large language model (LLM) or grammar-checking software specifically on the internal documentation (it will never leave the mainframe). Actually, the safest answer is the contrary of the current thinking in IT which would transfer everything to the cloud.

As a processor and system of record, the Hogan core banking platform has become the driving force behind some of the world's most influential banks. Now, we've enhanced Hogan's exceptional mainframe hybrid cloud solution by taking the best IBM Z solutions and providing blueprints for banks with aging cores to be "migrated to platform." Hogan x can even help you create a whole new digital bank. The platform is based on the principles of BIAN (banking industry architecture network) as being componentized and composable. It exploits Z Linux and other containerized and cloud-based solutions. The platform may also be offered in consumption based, as-a-Service models.

Find your best fit

Mainframe hybrid cloud architecture allows you to adopt a best-fit strategy for application implementation. The following is an example of an ecosystem in banking (between IBM Z and cloud). The applications are categorized into three separate areas:

• Digital channels client engagement
• Operational processing (e.g. order management, marketing and sales)
• Core transactions and data (core banking and credit cards)

These applications should be compatible with each other so that they share real time information and are able to maximize their business performance. Integration is vital for achieving wholesale interoperability. A mainframe hybrid cloud model with IBM zSystems helps clients optimize costs, performance and agility based on their application type and best-fit infrastructure. As an example, consider the digital channels. The high level of omnichannel interaction is based on the integration of partner and customer information to develop the optimal customer experience. This, and change of variable and workloads in cloud-based solutions, makes hybrid solutions your ideal solution.

Environmental, social and governance (ESG)

An often-overlooked advantage of the mainframe hybrid cloud architecture is environmental sustainability. Agile cloud working and energy-efficient mainframes will give banks the ability to maximize their data center footprint and data center energy usage (data centers consume approximately 1% of the global energy) which will help make the planet greener.

However, mainframe hybrid cloud architecture in banking is not just a technical choice. It is a strategic necessity, as it strikes the right balance between the protection, dependability and scalability and the flexibility and novelty required by digital working. With the combination of mainframe resilience and AI, ML and the flexibility of cloud computing, banks will be able to stay competitive, compliant and customer-centric. The success depends on the maximum utilization of this and such that the banks not only conquer the present issues but also are prepared and equipped to take advantage of the opportunities in the future. As technology evolves, AI and mainframe hybrid cloud strategies will play a pivotal role in banking's digital transformation, mapping a journey that pairs tradition with innovation, stability with agility and performance with efficiency. It is not time to be indecisive.

Someone in my office has just been to a major conference where one of the speakers mentioned this issue. He said: "We really do not know where technology is headed and so we will just sit back and wait to see where it emerges. The problem is that they will never get to see the point where it oozes since, in the exponential growth, technology will never stop, and at that point, it cannot be rationalized." The thing is that banks must board the train of foresight and open minds in order to implement AI as fast, safely and financially gainfully as possible.

The collaboration needs to be informed

That is why the assistance of the experienced providers is so precious. It is exemplary in the deployment of optical character recognition (OCR). In the case of one of our clients, we provided an option of four tools reading papers that said, it does not really matter which tool you choose. In 2 years, it'll be out of date." The procurement department also advised us to consider some others as well, which made it a full-scale exercise to attempt to identify the ideal tool. However, modern technology choice does not involve digging to the tool. It is about choosing and utilising a tool in the proper way, locating the correct business case and losing and gaining the technology. Technology has evolved to the point of being disposable, into planned obsolescence, and the next fix is only a couple of years before it becomes obsolete. In that way, instead of investing 5+ in a z16 solution, update your AI technology after every 2 years. That is, you should begin planning its replacement on signing the procurement contract.

Business issue: AI is used not always within the Z platform, where data has to move through networks. It is costly and not secure and it lowers your transaction-scoring capability in risk. Business impact: The IBM Z AI on-chip accelerator enables Hogan and related zLinux workloads to run real-time AI capabilities on the mainframe where the data currently exists. This enhances the performance of AI, security and cost. Furthermore, processes such as automated credit decision-making, and loan modification can be performed directly on the Z platform, and AI insights and business decisions can be faster and simpler to absorb. Celent projects that with 100 percent screening of all transactions, the individual banks would save about 100 million dollars.

Some of the possible applications of AI include credit scoring, fraud and AML. The mainframe platforms however run most of the banking and payment transactions, but the AI detection can be off-platform (less than 10% of transactions are run through an Al inferencing model in real-time because of latency, cost and customer friction problems). That is the reason why there are so many fraudulent transactions that go untraced and are not detected. Umbrella.ai can save you millions in fraud detection with on-mainframe inferencing. In this, Umbrella is the main element. The Umbrella application architecture is a proven technical platform that shall be integrated within the z/OS environment to be the quick integrated development environment (IDE) over 35 years. Umbrella has over 40 tier-1 financial institutions in various parts of the world and is used to execute approximately 100 applications by each bank. That is thousands of applications in mainframe development.

Umbrella.ai prevents fraud

Together with IBM, the duration of scoring or analysis of a transaction is being minimized. This will provide the clients with an opportunity to make virtual checks on 100% of their transactions and reduce time and cash wastage. Umbrella.ai goals include:

• 100 percent coverage of incoming transactions and current SLAs
• Utilizing existing models of frauds or implementing new models
• Decreasing scoring latency through the use of fraud prevention built-in on z/OS and the card authorization system
• Leveraging the MLz COBOL scoring service to greatly simplify the scoring call-out and reduce overhead to invoke the scoring service

The significant increase of fraud detection is one of the strong points of ML. Old rule-based systems are not flexible enough to address the emerging trends of fraud. Meanwhile, ML algorithms can keep updating themselves with the help of transactional data and detecting fraud cases more quickly and precisely.

Umbrella.ai and ML accelerate fraud detection and prevention

The AI/ML on IBM Z increases fraud prediction accuracy, allows actionable insights at a huge scale, and real-time fraud detection of fraudulent transactions is all possible. The higher precision of deep learning models allows them to reduce the high levels of false positives significantly. This implies that the banks have the ability to filter transactions in the process of detecting the fraud and still maintaining the customer experience and not bleeding money.

The partnership of Umbrella with AI and the IBM z16 delivers impressive business insights (no need for data science skills). Approximately 70 percent of financial operations are executed on IBM zSystems and that also is great news in terms of ESG the amount of energy used to infer on the mainframe is 40 percent lower than using a type of server farm.

Key capabilities include:

• Power any IBM z/OS application with Al-enhanced SQL and Umbrella.ai
• Discover and commercialize the hidden information within your data
• Determine similarities, differences and correlations
• Get interpretability with the box
• Use a single model on a set of questions
• Reduce the complexity of Al deployment
• Deep learning models can be used to improve false negatives

The IBM z16 supports the most popular ML algorithms, providing clients with an Al cloak to help them improve processes and drive greater business value from their existing investments.

Is it possible that you can afford to quail and watch your rivals establish a ravishing lead and sustainable advantage? Find out how AI and ML can be fully utilized to achieve the highest ROI. Also, learn how Hogan-powered mainframe hybrid cloud enablement can help you become a better fraud detector and prepare for future threats and opportunities.