While artificial intelligence (AI) has been around for several decades, the release of ChatGPT in late 2022 was a global gamechanger. It became the fastest adopted product of all time and catapulted AI into the everyday lives of consumers and businesses alike. Generative AI (GenAI) tools like ChatGPT and Perplexity are changing the way the world accesses, consumes and understands data – including financial and treasury data.
In the world of business-to-business (B2B) payments, there is increasing acknowledgement in the potential of AI or machine learning (ML) to deliver tangible value. When deployed in areas such as exception handling, transaction monitoring and fraud detection, AI is already being used to enhance operational efficiency, detect suspicious behaviour faster and improve the customer experience.
However, it’s important to cut through the hype – AI isn’t the solution for every problem. Finance teams should be pragmatic when deploying AI tools in the products and services they use.
The State of AI
A pragmatic approach requires, for one, a deep understanding of AI’s current limitations. Despite its rapid rate of development, GenAI, a core component of agentic AI, can still produce hallucinations and skew results. Even traditional AI can produce incorrect outputs due to data bias or faulty assumptions. Therefore, deploying agentic AI towards tasks such as executing large value payments might introduce significant risks at this early stage of commercial application. Human expertise and involvement remain essential in critical decisions.
Secondly, the “AI-first” wave of technology has created an assumption that AI is the key to removing any manual effort from day-to-day operations. In reality, implementing AI for simple jobs risks over-engineering. For basic tasks such as checking an account balance, AI adds unnecessary complexity and friction in the user experience.
The market also isn’t quite ready for agentic AI. Currently, in the absence of a widely accepted, standardised agentic payments protocol, giving AI agents banking credentials would likely breach banking agreements since only authorised signatories should make payments. The regulatory landscape is still evolving to address this, as well as more general concerns around data privacy, transparency, accountability, bias prevention, and risk management when deploying AI-enabled autonomous decision-making in financial services.
For the most part, even a trailblazing AI-ready organisation tends to rely on the symbiotic relationship between its workforce and AI assistants, rather than deploying fully autonomous AI agents to make all payments. Finance teams, therefore, should be focused on practical applications of AI that enhance human decision-making, not eliminate human intervention.
Practical Applications
AI has an important role to play in optimising corporate payments and cash management. It’s most suitable for tasks that require analysis of large data sets, are repetitive and rules-based, occur frequently, and have definable outcomes and a limited need for human judgement. The technology can shine in several critical finance areas, such as fraud detection, cash forecasting, payments optimisation, and investing surplus cash.
Fraud detection
Fraud detection is probably the most proven and widely-adopted use case for AI/ML in financial services. The technology has the ability to analyse high volume transactions in real time, recognise complex patterns and continuously adapt to new fraud methods, relieving fraud analysts of the manual burden of time-consuming investigations. The business benefit is early fraud intervention, reduced financial losses, regulatory adherence, and a better customer outcome due to reduced false positives.
Cash forecasting
AI is particularly useful for enhancing specific business processes, such as cash forecasting and reporting, as it can account for complex variables and adapt to changing conditions in real time.
In the context of cash optimisation, cash forecasting is a fundamental pillar on which to base forward-dated money movement decisions and underpins a strategic outlook. AI-powered predictive models can be used to analyse vast amounts of historical and real-time financial data, identify complex patterns and provide more accurate predictions of future cash flow compared to traditional methods.
Improving payment flows
Payments optimisation is another suitable AI application. B2B payments can be complex to navigate for multi-banked, multi-geography, multi-currency organisations, especially considering bank- and rail-specific cut-off times, fees and other costs.
A treasurer could receive AI-generated recommendations as to the most efficient and cost-effective route for processing payments. For example, they could split a high-value payment into multiple Faster Payments Service transactions, instead of sending a single payment through CHAPS.
In the context of a company that conducts payments in 20 countries, has 30 banks and more than 1,000 accounts, AI-generated payment flow recommendations could dramatically reduce costs and late fees, ensure payments are made and received faster, and allow finance teams to better plan their payment operations.
Surplus cash
CFOs can leverage AI to gain the greatest benefit from a company’s residual cash. From a CFO perspective, investing idle cash to deliver better yields is a strategic necessity in growing the business. However, it is a complex problem with many considerations, such as access to investment vehicles, market dynamics, regulations, cost, risk, etc.
AI can be used in automating the first two steps in the investment process – gathering information and analysing options – as well as supporting the decision-making step by presenting recommendations. At this critical point, a human should step in to select the optimal solution and complete the process.
This is a good example of human-AI interaction, and underscores the fact that AI should be used to facilitate, not override, key decision-makers in B2B payments.
The path to AI success
While the world isn’t totally ready for autonomous agents to perform critical tasks, AI, ML and GenAI are transforming the way finance teams access and consume information. Today, the technology is enhancing B2B payments and treasury workflows, helping teams gain clear answers to even relatively complex questions to enable better and faster decision-making.
In the current context, human involvement and intervention in critical decisions remain paramount. AI should provide options for humans to review, rather than make autonomous decisions.
Key best practices for implementing AI in B2B payments include identifying areas where AI can deliver the most immediate value, creating a framework and key performance indicators for evaluating AI use cases, and defining business objectives, such as increasing efficiency, reducing fraud or improving cash forecasting. Organisations should also assess their current technology stack and workforce readiness.
The final, and arguably most important, factor to ensuring successful in AI adoption is good-quality data. Finance and treasury teams need to ensure the data fuelling their AI models is clean, complete and consistent across all financial systems, to ensure AI in payments and fintech delivers reliable results. Robust data governance frameworks are critical, including implementing standardised data entry protocols and regular audits.
As in all technological innovations, market dynamics and timing are key factors of success. Given the current backdrop of ISO20022, which signals an increase in the quantity and quality of bank data, the timing of AI’s deployment more widely in corporate banking could not be better.
As a market leader in corporate financial data, AccessPay is well-positioned to serve high-quality, real-time, ISO20022-ready data, at a time when finance leaders are acknowledging that data is the key to unlocking the true value of AI.
We invite CFOs and treasurers to speak with AccessPay about their AI readiness.
