PYMNTS published a May 27 interview and video in its "What's Next In Payments" series, titled "When Legacy Becomes Leverage," featuring Garrett Baird, vice president of product, banking and FinTech at Paymentus. Baird told PYMNTS that "something like 75% of transactions are still running over mainframes," and argued that those legacy stacks are reliable and secure rather than simply liabilities. The piece frames AI as a way to orchestrate and extract more value from incumbent systems instead of replacing them outright. The coverage is presented as a recorded discussion; PYMNTS hosts the full video and a short article summary.
What happened
PYMNTS published a May 27 interview and video in its "What's Next In Payments" series, the May edition titled "When Legacy Becomes Leverage." In that discussion, Garrett Baird, vice president of product, banking and FinTech at Paymentus, told PYMNTS, "something like 75% of transactions are still running over mainframes," and noted the continued importance of reliability and security in payments infrastructure. The article frames the payments industry debate about "legacy" systems and presents AI as a tool to orchestrate existing stacks rather than an immediate one-for-one replacement.
Editorial analysis - technical context
Industry observers have increasingly treated AI as an orchestration layer that sits above heterogeneous back ends. Companies integrating AI with incumbent systems commonly focus on three technical tasks: surface-level automation (reducing manual workflows), data normalization for downstream models, and conversational or decision-support interfaces that wrap legacy APIs. These patterns reduce rewrite risk while still requiring engineering work to expose stable, auditable endpoints and to manage data quality for retrieval-augmented approaches.
Context and significance
The PYMNTS interview highlights a broader practitioner trade-off: mainframes and older stacks persist because they deliver reliability and security for high-volume transactions. For practitioners, that means many modernization projects are shifting from full rewrites to hybrid approaches that combine AI-driven orchestration with existing systems. This approach can change project scope-emphasizing API wrapping, robust logging, model evaluation against production data, and stronger governance controls.
What to watch
Observers will likely monitor adoption indicators such as increased investment in API layers on top of core banking systems, growth in tooling for productionizing LLMs with deterministic retrieval (RAG) over legacy data sources, and vendor integrations that explicitly target mainframe-to-LLM workflows. Also watch for reporting on latency, auditability, and compliance outcomes as organizations instrument AI atop transaction-critical systems.
Quoted material
The article records Garrett Baird saying, "something like 75% of transactions are still running over mainframes," and notes the payments sector debate over whether legacy is a liability or an asset to orchestrate.
Scoring Rationale #
Notable for payments and infrastructure practitioners because it reframes AI as an orchestration layer over existing systems rather than a simple replacement. The story is practical rather than frontier-level, so it ranks as a solid, actionable industry signal.
Practice with real Payments data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card