The Technology Isn’t New. The Economics Are.
AI and machine learning have been around for decades. Steve Mott from BetterBuyDesign deployed his first AI/ML application in 1990. What changed isn’t the intelligence; it’s the cost. Processing massive data sets in real time used to require serious capital. Now it’s affordable enough for mid-market merchants to put ML models into production.
Steve, along with Alx Block from Automattic and Joshua Lockhart from Aeropay, mapped the current state of AI in retail payments at the 2025 PaymentsEd Annual Forum. Their session covered where AI is showing real traction, where it’s still more promise than delivery, and how two very different companies are putting it to work.
A Framework for Where AI Fits
AI adds value in payment operations at three layers. At the tactical level, it augments existing processes by reducing fraud, friction, and errors, improving authorization rates, improving user communication, and automating data recording and access. At the strategic level, those tactical improvements compound into broader outcomes: better risk management, expanded sales, improved user experience, lower costs, and stronger customer retention.
The framework maps AI capabilities across the full transaction lifecycle: linking and identity (registration, stored profiles, security), shopping (sourcing, dynamic pricing, incentives), paying (checkout, payment method choice, rewards), and post-sale (fulfillment, returns, disputes). The starting point depends on your data and your pain points. A merchant whose biggest problem is chargeback representment needs a different AI approach than one focused on authorization rate optimization.
Automattic: AI for Chargebacks and Merchant Risk
Alx Block leads Payment and Risk Operations at Automattic, the company behind WordPress.com, WooCommerce, Tumblr, and Jetpack. Automattic powers over 43% of the web and processes payments through WooPayments in more than 100 countries. Two AI projects are in active development.
The first is chargeback representment automation. Automattic is building an internal tool that auto-fills and evaluates evidence packages, uses AI to match templates to reason codes, flags missing information, and automatically represents disputes. The goal is to reduce handling time and improve win rates at scale.
The second is a multi-agentic risk review system for merchant vetting. Multiple AI agents work in parallel: one checks website content, one verifies onboarding data, one scans for PII exposure, and one reviews social signals. The result is faster, more complete risk profiles. Analysts become leads of their own small team of agents rather than doing every check manually.
Alx also highlighted AI as a management tool: using structured frameworks for performance reviews, gaining visibility into trends across tickets and projects, and connecting individual contributions to business needs with less manual tracking.
Aeropay: ML for Account-to-Account Payments
Joshua Lockhart brought the perspective from Aeropay, which focuses on pay-by-bank transactions. Aeropay’s ML models are trained on billions of prior interactions and organized around four product pillars: bank linking (through open banking rails), payments (ACH, Same-Day ACH, RTP, FedNow), Guard (real-time risk and fraud controls), and Smart Recovery (automated retries with optimized cadences).
The Guard system uses predictive fraud and risk modeling that accesses account history, balances, and transaction behavior to provide merchants with guaranteed funds approvals in real time. The Smart Recovery engine uses user and bank behavior patterns to retry returned transactions with the highest likelihood of success.
The network effect matters here: every transaction strengthens the data models, which improves risk assessment and recovery, which leads to smarter approvals, fewer returns, and higher acceptance rates. It’s a flywheel that gets more accurate with scale.
What the Networks Are Doing
The networks are already moving. Adyen’s processor-outsourced risk management uses AI to handle fraud decisioning on behalf of merchants. Mastercard and PayPal are building facilitated checkout experiences powered by AI. And the emerging category of agentic commerce, where AI agents transact on behalf of buyers, is being explored by Visa, Mastercard, PayPal, and Stripe.
Agentic commerce is still early, but the concept is worth tracking: AI agents that can shop, compare prices, and complete purchases on a consumer’s behalf, with the payment networks providing the trust and authentication layer.
Getting Started
The AI/ML opportunity in retail payments is real and growing. Constructive paths exist for automation across the spectrum of payment operations. But sorting out what works for a given merchant takes diligence. Start with your data and your pain points, not with the technology.
This session was presented by Steve Mott, Principal at BetterBuyDesign, Alx Block, Head of Payment and Risk Operations at Automattic, and Joshua Lockhart, CTO at Aeropay, at the 2025 PaymentsEd Annual Forum.
If AI in payments is on your radar, the PaymentsEd Forum is where practitioners share what’s actually working in production. Save the date for the 2027 Forum.