Predictive payments: How AI can enable new payment and cash management services

The banking industry is rapidly adopting artificial intelligence (AI) to enhance the customer experience, increase operational efficiency, and gain a competitive edge. Chatbots powered by generative AI (GenAI) are becoming increasingly common, improving customer support and resolving queries 24/7. AI algorithms are optimizing real-time fraud detection and customer behavior analysis, as well as supporting hyper-personalized product recommendations.

At the recent North American Payments Summit, Tracy Lagasse, Microsoft’s innovation lead for financial services institutions, commented that, “AI’s impact on banking is as comparable to the invention of the printing press” in its seismic impact on the sector.

With payments at the heart of a bank’s business, it’s no surprise that banks and solution vendors alike are assessing how AI can enhance payment processing. Improved fraud detection is held up as the crown jewel in applying rules-based AI and machine learning (ML), but now banks are exploring how AI can improve straight-through processing (STP) rates (e.g., automatic payment repair based on historical actions) and assist customers with payment origination.

Driving new payment capabilities through AI analysis

Banks hold significant amounts of data about their customers’ payment histories, including individual retail banking customers, small- and medium-size enterprises (SMEs), and large corporates.

For individual customers, AI data analysis can reveal payment trends, such as who is making payments, when, and for how much, enabling banks to create predictive personalized reminders as to when to make payments (including automatic payments pre-approved by customers). This could extend to an analysis of customer balances and inbound cash flow (with customer approval) for the purpose of suggesting more efficient cash uses, such as redirecting cash to savings or investments.

For SMEs and large corporates, particularly those that are single-banked, AI analysis of historical payment data can reveal patterns in accounts payable/account receivable payments. Smaller businesses may not have sophisticated ERP systems to manage payments efficiently. Thus, banks can step in with AI tooling to offer similar capabilities.

For SMEs, as well as single-proprietor businesses, AI analysis of historical patterns of account receivables (assuming visibility of the payment request/invoice) can enable the bank to suggest novel payment terms that reduce days sales outstanding (DSO) or offer better credit lines with less risk for the bank. Though not specifically AI-related, more rapid implementation of Request to Pay schemes could dramatically improve the account receivable process.

For accounts payable, supplier payments can be optimized (held) to be executed just in time, leveraging instant payment rails based on history (or visibility of invoicing due date) to optimize customer balances.

Technical challenges and considerations

As with any new technology, testing is required to identify the right AI solution, along with an ROI analysis. Although large language model (LLM) GenAI is relatively new, ML and rules-based AI tools and solutions are well established. As a result, identifying appropriate AI solutions for use cases and data analysis should be relatively straightforward.

Data considerations in choosing AI software include the following:

• Quantity: How much data is required for the analysis? For example, how many years of payment history is needed or is other customer information required (e.g., from a card payments database or from access to a customer’s payment data held elsewhere, using open banking APIs).
• Quality: The use of richer ISO 20022 payment data, where possible, can lead to better insights.
• AI training: The application of appropriate test data sets to tune the AI algorithm is essential.

Responsible AI use requires human oversight and AI output reviews to explain results and ensure AI decision-making is not purely “black box.” It’s important to note that traditional algorithmic data modelling and analysis can lead to “unusual” results. Explainable AI (XAI) solutions may be considered when requiring customers to provide the reasons behind a particular decision.

Regulatory challenges and considerations

However, despite the promise of AI, the regulated nature of banking means that applicable laws and regulations must be carefully considered when applying AI technology. These include new AI-focused regulations such as the European Union AI Act now in play (from July 2024), as well as existing and impending laws such as the General Data Protection Regulation (GDPR) and the Digital Operational Resilience Act (DORA) (in particular DORA’s right to be forgotten and its governance of customer data in AI-powered products). Consideration also must be given to solution auditing, particularly LLM-based solutions, to ensure adherence to applicable data protection regulations.

More immediately, AI services will require customer opt-in, potentially requiring customers to adhere to updated terms and conditions related to data use.

In conclusion, AI tooling applied to payments holds much potential in generating new value-add services for bank customers—from individuals, to SMEs, to large corporates. If you’d like to learn more about how the responsible use of AI can generate valuable payment outcomes, feel free to contact me. Also, learn more about CGI’s AI expertise and capabilities in general, as well as our AI in banking thought leadership and solutions.