How to use AI in payments to optimize, protect and grow

By effectively implementing AI in payments, financial institutions can drive customer engagement and achieve measurable growth.

In brief

• The use of AI in payments is rapidly expanding, providing the opportunity for financial institutions to reach new markets and boost revenue.
• Organizations should use key performance indicators (KPIs) to measure how AI impacts revenue, risks and operational efficiency.
• A structured approach to deploying AI is essential for organizations to gain the full benefits and improve customer experiences.

Tracking AI’s impact on payments 

To turn AI into a tool for measurable growth, organizations need to connect it clearly with their business goals using specific KPIs. These KPIs help track AI’s impact on revenue generation, risk reduction and operational efficiency. 

KPIs to track include the following:

1. Optimize

Payment processing cost per transaction, dispute resolution cycle time, exception handling automation rate and application performance interface uptime

2. Protect

Fraud loss rate, false positive rate, compliance adherence and AML alert-to-case conversion rate

3. Growth

Revenue from new payment products, digital channel adoption rate, customer activation and engagement rate, and revenue per transaction channel

Leveraging AI in payments

Enhancing payment capabilities begins with a systematic mapping of front-, middle- and back-office functions to match the payment process. This clarity helps operations run more smoothly and highlights areas that need improvement. Once the mapping is done, organizations can better evaluate their current abilities, find gaps and create focused plans to improve their payment systems.

Significant developments in payments are expected over the next two years. Examples of market trends and use cases include:

1. Optimize

• Reduced fees: AI-powered smart routing can lead to cost savings of up to 26% on US debit transactions.⁴ It optimizes authorization rates and processing fees using real‑time data such as card type, location and historical success rates.
• Automated reconciliation: AI models streamline back-office operations, reducing manual review efforts by 70%.⁵

2. Protect

• Enhanced fraud detection: AI analytics can address over $40 billion in annual fraud losses, leveraging 12 months of transaction history in real time to identify fraud patterns early before they lead to a financial loss.⁶
• Improved verification: AI models enhance verification accuracy, cutting account validation rejection rates by 15%-20%.⁷ This lowers payment exceptions, returns and misdirected payments.
• Faster fraud response: AI-based filtering increases conversion from alerts to actionable fraud cases by 30%-50%.⁸

3. Grow

• Personalized offers: AI-driven personalization can lead to over 30% increases in redemption rates for credit card offers and merchant funded rewards.
• Market expansion: AI-based onboarding reduces merchant approval time by 70%.⁹

Three key use cases demonstrate how AI-driven solutions can provide measurable benefits 

Implementing AI across the enterprise

Effectively integrating AI calls for a playbook with a well-defined, repeatable strategy. The following journey is divided into four main phases: 

1. Backlog build

Compile a broad list of possible AI use cases across operations. Rank according to desirability, viability and practicality. After prioritization, the AI governance council reviews and approves the final selections. 

2. Scoping and design

Clarify the end users and expected outcomes. Gather relevant data sources, outline the data pipeline and craft a blueprint for the AI solution. Define the deployment strategy by identifying key data elements and orchestrating required data sources, while working to ensure alignment with AI Center of Excellence (CoE) standards.

3. Build solution

Develop the necessary data and infrastructure pipelines, conduct feature engineering to ready the data for AI models and create a plan to evaluate performance. 

4. Test and deploy

Conduct thorough model testing and validation, perform user acceptance testing to identify performance issues and unexpected behaviors, and secure all necessary approvals before moving the AI solution to production. 

For successful deployment, organizations should also establish solid foundations — such as robust data infrastructure, access to adequate computing resources (e.g., processing power, storage, scalable environments) and a deployment framework that governs testing, release, monitoring and ongoing model management by following industry-leading practices. In addition, focusing on clearly defined business objectives, thorough implementation plans, strong talent management and effective governance are critical to success. 

Adhering to this structured method and prioritizing foundational readiness allows organizations to unlock the benefits of AI, fostering innovation, improving customer experiences and delivering tangible business results.

Thank you to Barjinder Badwal, Senior Manager, Technology Consulting, Ernst & Young LLP; Anant Gupta, Manager, Technology Consulting, Ernst & Young LLP; and Akshaykumar Pole, Senior, Technology Consulting, Ernst & Young LLP, for contributing to this article.