Artificial intelligence can dramatically improve the speed of transactions and the accuracy of anomaly detection. To help realise the full potential of this technology, we’re building a state-of-the-art AI platform and innovating with our community.
AI algorithms can rapidly analyse large quantities of data, learning through repeated use and constantly refining their results. When it comes to tasks like anomaly detection, this technology is far more effective than today’s rules-based systems, and more efficient than a human could ever be.
These characteristics mean AI algorithms are also well suited to monitoring networks for outages, screening payment instructions for missing data, and identifying fraudulent or sanctioned transactions. With this technology at our fingertips, the financial community has the potential to transform the way it analyses data and process transactions.
“AI and machine learning aren’t new technologies in themselves – their building blocks have been around for quite a few years. The real trick is coming up with new creative ways to train these algorithms – improving their performance, making them more precise and using them to transform the way our entire industry operates.” says Thomas Zschach, SWIFT’s Chief Innovation Officer. “We want to uncover the biggest opportunities and hone them so that institutions of all sizes can benefit.”
Partnering to build an enterprise-scale AI platform
Machine-learning models need large amounts of data and demand a huge amount of computer memory. With 11,000 members sending millions of transactions across our network every day, we recognise the need for a highly scalable architecture without data-resource limitations. That’s why we’re building a high-performance, industry-leading AI platform.
Collaboration has always been at our core. And when it comes to innovation, working with others also offers the best route to success. “These partnerships are incredibly valuable,” explains Chalapathy Neti, Head, AI and Machine Learning Platform at SWIFT. “Each member of the partnership can concentrate on what they do best, which ultimately leads to us achieving more than we could ever do alone.”
This state-of-the-art AI platform embodies the value that good partnerships can deliver. It’s built on C3.ai’s AI platform for model development and deployment, with Red Hat OpenShift’s container platform on hand to build and scale applications. Using unique software-defined memory, provided by Kove, there’s no data-processing limits – just scalable memory, on demand.
We’ll be launching later this year. Once live, we’ll be able to start using this creation to its full potential – enhancing our existing products and co-creating new ones with our customers and partners.
A foundational model for anomaly detection
One of our early objectives is to develop a stronger model for anomaly detection. This will demonstrate the value that AI can deliver, enhance today’s rule-based systems and bolster existing our products like Payment Controls and Payment Pre-validation. The result – a secure and frictionless payments experience for our entire community.
By applying machine learning algorithms to these tasks, we can enable more precise analysis – fewer false positives to investigate and fewer rejected messages to repair. In the future, AI could even enable automatic correction of errors on input, optimise routing options for payments and streamline processing in many other ways.
To try and enrich our AI model, we’re listening to our community – understanding which issues occur most often within the payments process. We’ll also use the model to co-create AI-native services with customers and partners and to develop other use cases for AI.
Exploring the potential of federated AI
Within finance, the race is on for institutions to stay one step ahead of fraudsters who threaten to compromise the security of payments networks around the world. Today, these attacks aren’t just limited to a single country or company – they often have a much wider scope and, as a result, have the potential to do much more damage. To keep pace with these developments, it’s essential that our industry continues to develop advanced techniques and technologies that can keep finance secure.
And federated AI might just be the answer. It’s a machine learning technique that’s able to transcend institutional borders and analyse data across multiple organisations, generating much richer insights than a single institution could generate alone. By doing this, machine learning algorithms are exposed to a much more diverse range of data, making them more effective at spotting fraudulent transactions in the real world.
“While federated AI has a lot of potential within finance, it’s important that the data used to train algorithms is kept anonymous, and that sensitive information is kept confidential,” continues Neti. “To help keep our community safe from fraudsters, we’re working to realise the full potential of this technology – exploring the potential of federating data from participating member institutions using privacy-preserving secure AI networks.”
It’s clear that outside the lab AI has the potential to deliver solutions for a number of the challenges our industry faces. These initiatives form the core of our AI innovation plans for 2022, and with them and others to come, we look forward to working with our community to uncover the full potential of AI. If you think your institution could help transform the future of finance, we want to hear from you. Get in touch at firstname.lastname@example.org.