By Stuart Tarmy, Global Director, Financial Services Industry Solutions at Aerospike

Data has the power to set organisations apart. Leveraging it to enhance real-time applications from customer experience to fraud prevention, instant payments and reducing churn. The more data-driven an organisation is, the better it performs.

But data set sizes are growing at a phenomenal rate and real-time data is growing even faster. The financial services industry is deploying AI to help manage this explosive growth. While this is in essence the right way to progress, organisations need to move forward with knowledge and caution. The road to deploying AI and machine learning has many nuances, and there has been much hype around what AI can and cannot do. It is critical for companies to understand the most productive use cases, how to best leverage AI and advanced AI algorithms such as neural nets, and the underlying technical architecture to make these work in real-time to maximise the customer experience.

A McKinsey study conducted a couple of years ago found that during the pandemic most high-performing companies boosted their investment in AI, with the financial services sector increasing by 28 percent. This was further reinforced last year by research from NVIDIA which showed that 80% of the world’s top financial firms are spending billions on AI to improve services and sharpen their competitive claws.

The message is clear: embrace AI or you will fall behind. Both sets of research, and other studies, indicate that it is the ‘top’ financial services firms that are spending money to deploy AI solutions. What about everyone else? Should we assume that without AI they will fall behind in their growth, their ability to serve customers satisfactorily and perform for shareholders?

The situation is not that simple, because there are multiple fronts that all companies should address if they want to truly leverage the power of AI.

Improve model development and performance

To ensure AI systems work to their full capacity they need great data scientists with the ability to build sophisticated AI algorithms. They should be able to handle large volumes of data typically measured in terabytes, petabytes, and sometimes even exabytes in some industries, because the more data you have, the better you can train the model and determine the critical data attributes.

In typical financial services use cases, business units and customers are expecting a real-time experience – something in the region of sub-milliseconds or the blink of any eye – and that requires a super-fast engine. This is provided by multi-model real-time data platforms that not only deliver low latency but can also provide predictable performance at any scale.

These platforms serve as a system of record and can handle transactions in their millions regardless of whether they take place on mobile phones, tablets, laptops, at ATMs or in branches. They can help companies to model risk and provide analysis on the market, credit and liquidity exposure across multiple asset classes and customers. Added to that are features for providing identity management, detecting fraud, personalisation and compliance – all essential in financial services. The best-in-class real-time data platforms are multi-model, meaning they can provide multiple capabilities such as NoSQL, graph, time-series and JSON document functionalities.

Introduce large, multiple types of data to improve AI system performance

Look around at some of the world’s most advanced technology users, and we can see that they have adopted neural nets – which work on the concept of neurons talking to each other inside a computer system – and deep learning systems. This gives them the power to analyse millions of data attributes to train a model, sometimes as many as ten million attributes. At PayPal, for example, they have seen improvements in performance of 30 percent by deploying neural nets over more traditional AI/ML systems.

Another key industry trend is incorporating Explainable AI into the system design. This provides explanations about why the model is reaching certain conclusions, and the decision making it did to arrive there. It allows the system to be tweaked and adjustments to be made so that the model works more efficiently. Companies can access fast and reasonable explanations about how their data will be used which is especially useful given data privacy laws such as GDPR and CCPA, as well as the newer AI regulations that have been proposed, such as the EU’s AI Act.

Building and deploying AI for real-time performance

Let’s go back to real-time, multi-model data platforms and look at another way they support financial services companies.

We would expect them to enable fast model testing, but they can also ingest data extremely fast from a variety of sources at the edge. This helps systems of engagement to act in real-time, delivering stellar experiences to customers in line with their performance expectations and quickly serving the needs of employees.  This level of speed and functionality is essential, particularly in consumer lending, payments, credit cards and fraud prevention; for enterprise platforms that are being used for analytics, customer360 and personalisation, and to support systems that protect against cyberattacks.

These highly functional real-time data platforms can sound very expensive, but the newer ones use more modern architectures to minimise costs and reduce total cost of ownership (TCO).  It’s not unusual for these more modern data platforms to reduce TCO by up to 80 percent or more.  They do this by significantly reducing the size of the server footprint needed and achieve best-in-class real-time speeds by techniques such as working natively with the underlying hardware and incorporating sophisticated data parallelization and memory management techniques.

The benefits of AI in a multi-cloud, on premises hybrid environment

Many financial services firms, while moving much of their infrastructure to the cloud, still have legacy systems and are currently running in a hybrid environment. They require a real-time data platform that can run in this type of hybrid environment.

Multi-model, real-time data platforms can be deployed as software, regardless of whether the user is multi-cloud, on premises, using a single cloud provider or all three. It’s always on, so its performance is consistent throughout the database along with multi-site clustering that allows companies to put clusters in different data centres without sacrificing consistency.

So, if you are looking to embrace AI or neural nets to thrive and dominate in today’s competitive market, you need an engine that is powerful enough to analyse large amounts of data in real-time. Companies across the world are already enjoying savings that run into millions of pounds by reducing their server footprint and improving their real-time transaction throughput. Most importantly, they are gaining marketshare and meeting customer expectations that allow them to remain at the forefront in an increasingly competitive market.

The post Demolish Barriers to Leverage AI in Financial Services appeared first on Finance Derivative.