Artificial intelligence (AI) is moving from experimentation to practical deployment across financial services. As agentic models exceed merely generating recommendations, issuers are discovering that AI itself is no longer the limiting factor for autonomous action.
AI readiness varies significantly across issuer segments. Digital banks, FinTech issuers and institutions operating modern cloud-native architectures often have greater flexibility to connect workflows and deploy new capabilities rapidly. Visa’s research finds that 60% of neobanks and FinTechs report high or transformational improvements from AI deployment in transaction risk management, compared with 44% among incumbent banks and 42% among local and savings banks.
The gap is not necessarily a reflection of AI ambition. Rather, it highlights differences in operational complexity, technology architecture and workflow connectivity. Large, incumbent issuers often manage decades of accumulated technology, multiple processors and highly fragmented operational environments.
As a result, the challenge is not universal AI adoption but operational transformation at different levels of complexity. The common requirement across segments remains the same: the ability to operate a modern card-issuing platform with connected data, systems and workflows in ways that allow AI-driven processes to execute reliably.
AI Is Advancing Faster Than Issuer Operations
Issuers are caught between experimentation and scale, particularly incumbent, local and savings banks with fragmented systems and payment data.
Most institutions remain caught between experimentation and scale.
AI’s value is becoming increasingly clear. Organizations continue to grow their AI investments because measurable business outcomes are already emerging. NVIDIA reports that 89% of FIs say AI has increased revenue or reduced costs, with operational efficiency (52%), employee productivity (48%) and customer experience (37%) among the leading benefits. Nevertheless, after several years of experimentation, banks’ enterprise deployment remains limited. EY reports that 52% of banks have piloted agentic AI, while only 16% have fully deployed production use cases.
52%
of banks have piloted agentic AI, while only 16% have fully deployed production use cases.
The reason is becoming equally clear: Scaling AI’s value across an enterprise requires something many organizations have yet to build.
Readiness depends on operational maturity, not AI maturity.
Beyond model intelligence, AI in issuing systems depends on the ability to access reliable data, connect workflows and operate within clear controls. The organizations pulling ahead are distinguished by the maturity of the environments supporting them.
New research from Visa reinforces this divide. Although 75% of European banks have a defined AI strategy, just 43% report an organization-wide approach to identifying and testing AI use cases. That process requires coordinating data, workflows, governance and operational execution across platforms that have historically operated in silos.
The challenge is creating the operational environment where AI can reliably execute work while maintaining security, regulatory compliance and customer trust. That distinction explains why many AI initiatives stall.
Agentic Operations Expose the Limits of Legacy Issuing
As AI begins to participate directly in operational workflows, issuers face growing pressure to modernize architectures designed for human collaboration that are now being asked to execute work across systems.
Architectures built for coordination are being asked to support execution.
Legacy architecture, fragmented workflows and inconsistent data remain among the greatest barriers to AI adoption in issuance. In Forrester’s global survey of banks and issuers, 50% identified legacy payments infrastructure and technical debt as key modernization challenges.1 Unlike a general-purpose consumer tool, operational AI must draw upon an institution’s own transaction histories, customer records, risk signals, policies and real-time account information to make accurate decisions and take appropriate action. FII International identifies data quality, legacy systems and talent shortages as the principal constraints preventing organizations from scaling AI successfully.
For incumbent banks in particular, Visa finds that fragmented technology estates and legacy architecture can constrain AI’s ability to operate within core workflows rather than alongside them.
The impact of fragmented operations is already visible across the payments ecosystem. AutoRek reports that 80% of organizations experience moderate to severe operational disruption resulting from fragmented payments data. The finding illustrates a broader challenge for issuers. When information, processes and operational actions remain distributed across disconnected systems, organizations struggle to execute consistently and efficiently across the card lifecycle.
Manual operational processes further slow execution. Reconciliation, disputes, chargebacks and card renewal remain heavily dependent upon human intervention at many institutions, even though automation has demonstrated dramatic improvements in speed, consistency and customer experience.
Removing individual bottlenecks helps, but AI demands something larger than isolated modernization projects.
Modern Issuing Platforms Enable Agentic Operations
The next generation of operational efficiency will be built on platforms that provide the real-time visibility, orchestration and governed execution capabilities required for agent-driven workflows to operate across the card lifecycle.
Real-time issuing infrastructure enables operational execution.
For AI to move beyond generating recommendations and begin taking action across issuing workflows, operational systems must function as a connected ecosystem. Application programming interfaces (APIs), workflow orchestration and interoperable data provide the connective tissue that allows AI to coordinate activity across onboarding, servicing, authorization decisioning, transaction processing, fraud management, card and credential lifecycle management, and customer support.
Modern issuing platforms address this execution gap by turning operational processes into connected, API-driven workflows. APIs allow systems to interact in real time, make data accessible across environments and execute actions consistently. This transforms static workflows into dynamic execution layers where agentic systems can operate effectively, under clear rules and controls.
In practical terms, this means an issuer should be able to move from a customer request to an operational action without waiting for batch-file exchanges or manual coordination between teams. APIs are not just a technical upgrade in this narrative. They are the mechanism that allows agentic operations to move from “AI-assisted decisioning” to “AI-enabled execution.”
Card replacement shows what real-time orchestration looks like in practice.
Automated card lifecycle management demonstrates how operational modernization creates measurable operational benefits.
Card replacement is one of the clearest examples of why issuing modernization matters. A single lost, stolen or damaged card request can involve multiple dependent actions: physical card issuance, virtual card issuance, address verification, PIN management, token updates and coordination with external partners such as personalization bureaus.
In many current environments, this urgent request depends on disconnected steps with multiple manual handoffs. This process slows execution and increases operational complexity.
In a modern, orchestrated model, the same card replacement request becomes a connected lifecycle event. Once the request is validated, the issuing platform can trigger the relevant actions in sequence or in parallel: confirm the customer’s details, issue or replace the physical card, provision or update the virtual card, manage PIN-related steps, refresh associated tokens and coordinate fulfillment with the personalization bureau. This is the shift from operational handoffs to real-time orchestration.
For example, in many issuing environments, cardholders receive limited visibility into the status of a newly issued or replacement card. When delays occur, customers often contact the bank for updates. Industry observations show that approximately 25% of inbound contact-center calls relate to card status inquiries such as “Where is my card?”2 By connecting issuance, fulfillment and customer communications through a modern issuing platform, issuers can provide real-time status updates and proactive notifications, reducing operational costs while improving customer experience.
25%
of inbound contact-center calls relate to card status inquiries such as “Where is my card?”2
Modern issuing platforms orchestrate these activities through a unified workflow, enabling autonomous agents to coordinate card reissuance, update payment tokens in real time and trigger integrated customer notifications automatically.
The barrier is not whether an AI-assisted or agent-driven workflow can identify the next best action. The barrier is whether issuing infrastructure can execute that action reliably across systems, partners and channels.
The same orchestration challenges appear in areas such as card product configuration and transaction disputes, where execution depends on coordinating actions across multiple systems and teams. Whether launching a new card product or resolving a disputed transaction, issuers must synchronize data, rules and workflows across authorization, tokenization, customer servicing and operational platforms.
The future belongs to issuers that redesign operations, not simply add AI.
Competitive advantage will belong to issuers that modernize their operations with real-time processes in every system rather than attempting to layer AI onto yesterday’s operating models. The potential upside is substantial. McKinsey estimates that a future operating model involving AI agents could reduce overall bank costs by approximately 25%, translating to $250 million to $500 million in bottom-line impact for every $100 billion in assets. Much of that value would come from transforming middle- and back-office work rather than merely adding intelligence at the customer interface.
Building AI-Ready Issuing Operations
AI transforms prepared organizations, not merely organizations that adopt AI. The next generation of leading issuers will therefore be defined not simply by AI adoption speed but by their ability to combine execution, agility and scale with durable operational trust.
PYMNTS Intelligence recommends the following actionable roadmap for organizations preparing for the next phase of AI adoption:
- Assess operational readiness, not just AI capabilities.
- Identify workflow fragmentation and systems that limit execution.
- Modernize these workflows and systems with a card-issuing platform using APIs and orchestration to connect AI with the systems, data and workflows required for execution.
- Prioritize governed, real-time issuing data that enables AI to act consistently across the bank’s operations.
- Build governance and human oversight alongside AI expansion.
In the race to deploy AI, modernizing the operational foundation is already underway. The institutions that finish building the road first may ultimately travel the farthest.
1. Forrester. The State of Payment Tokenization: Ambition, Complexity, and the Path Forward. A Forrester Consulting Thought Leadership Paper Commissioned by Thales, June 2026.↩
2. Per Thales’ discussions with customers.↩
