This is an excerpt from the Future of Payments 2025 report.
Never has technology evolved as quickly as it does today, and innovation is ripe in finance and banking. AI, tokenisation, stablecoins, CBDCs and hyper-personalisation are all elements of discussion when it comes to making transactions more secure, streamline
operations, reduce costs and enhance customer experience.
Let’s explore these trends more closely.
AI in banking and payments
Payment service providers (PSPs) have used rules-based AI, such as machine learning and robotics process automation, for a long time. Now, new forms of AI – centred around large language models (LLMs) or generative AI, enable new use cases and offer institutions
innovative opportunities to differentiate.
One of the main use cases lies in AI’s fraud detection and mitigation capabilities. Many organisations have started deploying AI-based AML and fraud solutions that help them to improve their detection rates by reducing false negatives or positives and reducing
the manual labour needed to mitigate modern threats.
The increased availability of generative AI can additionally enable financial institutions to reduce the time needed to investigate potential suspicious activities and has the potential improve the Know Your Customer (KYC) processes.
Amelia Ruiz Heras, head of global solutions consulting, payments at Finastra, highlighted: “The technology can provide automatic sourcing and validation of customer information, flagging anomalies to the account manager, and assessing client-specific market
data and media stories to ensure that KYC criteria continue to be met. With other functionality such as voice-activated payment authentication, improved counterparty screening and cross-checking more data records in shorter timeframes, the technology has several
promising applications for robust payments security.”
Synthetic data is another opportunity unlocked by AI. In 2025, data is one of the most valuable assets in any organisation, and institutions are not just faced with a vast amount of data they have access to and need to safeguard, but complex, comprehensive
data is also needed to train their fraud prevention models. In order to protect sensitive customer data, synthetic data is one of the most effective ways to address the data question.
“AI-generated synthetic data is also being explored to provide institutions with more comprehensive and realistic business and market outlooks, thereby improving cashflow forecasting accuracy. This data could also be used to detect new types of fraud, ensuring
institutions stay ahead of fraudsters as their methods continue to evolve,” Ruiz Heras added.
If we look at the back-office, AI has the potential to improve operations, strengthen security and boost productivity, which can help organisations drive down costs significantly. On the flip side, in the front-office AI can drive revenues by unlocking new
revenue streams and allow access to previously untapped potential.
If we look at the front-office, hyper-personalisation is one of the key opportunities unlocked by AI. Hyper-personalisation enables organisations to address increasingly complex customer expectations by offering highly tailored products, streamline the customer
service journey, and also improve security.
Sulabh Agarwal, global payments lead, Accenture, noted the importance of ISO 20022 in these efforts, he stated that “its enhanced data formats allow for the exchange of richer information such as structured remittance data and purpose codes. This improves
straight-through processing (STP) rates, reduces errors, and offers greater insights for businesses. This, in turn, reduces fraud and increases the efficiency of global payment systems and allows the businesses to establish better experiences for the end consumers.”
In collaboration with the increased data that will be available with ISO 20022, organisations will now have a richness of information on their customers across a wide ranging financial landscape. The available data, in combination with AI and behavioural
analytics, which will help combat fraud by knowing what a customer’s normal and thus abnormal transactions look like.
Yet there are some hurdles to overcome in order to achieve this, as Kevin Flood, director, FIS payments ecosystem strategy, corporate and international banking, FIS Global pointed out: “The risks for what can still be considered a fledgling technology, and
one that is yet to be fully regulated, remain high. The impact of recent AI regulation will take some time to unpack and implement, which means that risk appetites remain somewhat on the more cautious side.
Nevertheless, the use cases for AI continue to grow and the relative certainty of regulation will allow for the envelope to be pushed. The use for AI and GenAI for hyper-personalised products should take the charge in conjunction with using it to fight the
unabating rise of fraud.”
Looking to the fraud risks, Agarwal noted there are a number of different approaches being taken globally, such as Confirmation of Payee in the UK, fraud detection at multiple stages in Sweden, the launch of Transaction Monitoring Netherlands, and Malaysia’s
recent National Fraud Portal. Yet he argued that “an ideal approach would involve all the industry stakeholders (payments market infrastructure, banks, telcos, social platforms, law enforcement agencies, regulators) and focused on the entire value chain of
fraud prevention, fraud reporting, collective action, fraud recovery and customer reimbursement. Cross-industry collaboration through shared fraud databases and real-time information exchanges, allows organisations to stay ahead of the new techniques deployed
by the fraudsters.”
Considering AI technology is in its nascent stages, Ruiz Heras outlined four steps that institutions should take before making a decision on whether or not to deploy a solution:
- Decide how you will gauge the benefits AI will bring, such as reducing TCO and losses to fraud, while increasing operational efficiencies and straight-through processing (STP).
- Pair these metrics with softer benefits such as enhanced client experiences, having the means to innovate more easily, building your brand through improved and AI-generated materials.
- Explore with technology partners any alternative pathways to your objective. Evaluate both the nominal investment associated with each pathway as well as the level of disruption and risk each pathway will impose.
- Consider the risks of rapidly evolving technology, keeping in mind that some of the most profitable applications of AI are perhaps yet to be unearthed. Ask your technology partners to keep you appraised of emerging use cases. Ensure you understand AI usage
policies and processes your partner has established in their organization, as well as any regulatory requirements you might need to adhere to, such as the EU AI Act.
Challenges and benefits of CBDCs, tokenised deposits and stablecoin adoption
Tokenised money is on the horizon in many regions across the world. As of September 2024, 134 countries and currency unions, representing 98% of global GDP, are exploring CBDCs. Three CBDC schemes have been launched so far (in the Bahamas, Jamaica, and Nigeria),
and 44 – including most of Europe and APAC) are in its pilot phases. North American CBDC schemes are still in development.
While development marches relentlessly on, Annalisa Ludwinski, head of correspondent network management, Investec, noted the challenges of the situation: “Market fragmentation is already a concern, as many institutions are researching and developing use
cases for tokenised money and assets at a pace that regulators and central banks cannot keep up with. There are already costs and challenges associated with connecting to multiple private loop networks, with little to no interoperability between them; this
issue will only increase.”
Naveen Mallela, co-head of Onyx by J.P. Morgan, added: “Market fragmentation is inevitable as different parties develop and test out different innovative solutions. The alternating cycles of divergence and then convergence is a natural part of innovation.
While we will be better off with better solutions, the key is to drive convergence quickly and minimize the period of fragmentation.”
One concern is the programmability of tokenised assets. Petia Niederlaender, director payments, risk management and financial literacy, Oesterreichische Nationalbank, highlighted that programmability raises privacy concerns, might lead to new operational
and cybersecurity risks, and could potentially have a negative impact on economic stability as well as create new challenges for regulators.
“In this context, it is key to distinguish between ‘programmable money’ and ‘programmable or smart payments’. Whereas ‘programmable money’ contains embedded rules that determine how the money can actually be spent, “programmable or smart payments” are conditional
payments that are carried out automatically once a certain pre-defined set of conditions has been met,” Niederlaender explains.
Yet, she added, this is also where innovation plays a crucial role and where the digital Euro could potentially turn into a major driver of future advancements: “As a key payments’ infrastructure, the digital euro would provide the basic railings upon which
fintech companies and start-ups could innovate and build new use cases, value-added services and convenient consumer and business solutions. While the digital euro’s core functionalities would ensure broad public access and inclusion of vulnerable groups,
private-sector market participants could leverage this by offering their enhanced services and valueadded solutions including conditional payments via APIs.”
Increased efficiency is not the only benefit that can be realised from programmability. Tokenisation additionally offers greater access to financial services, especially for un- and underbanked groups, which boosts democratisation in the space.
Lastly, Ludwinksi added positive effects on global trade linked to programmability: “If we consider the development of a stablecoin to help smooth global trade, there are certainly benefits. If stablecoins could be linked to a smart contract that would,
for example, automatically pay out using that stablecoin when certain parameters are triggered, then the operational efficiency savings could be substantial, as well as providing an excellent way to manage risk and exposure on an almost real-time basis."