Matt Pearce, Vice President, Fraud Risk Management and Dispute Operations at i2c, spoke with Fintech Business Asia about how AI and real-time decisioning are helping financial institutions combat increasingly sophisticated fraud while reducing friction for legitimate customers. He also shared his perspective on how AI could reshape credit, disputes, personalization and other areas of digital financial services. .

Q: i2c has been named the Most Innovative Fintech in Latin America by Global Finance, with its AI-driven Fraud Risk Management solution highlighted. What specific challenges in the Latin American payments market led i2c to develop this solution?

A: The payments landscape across Latin America is evolving quickly. As digital payments adoption accelerates, financial institutions and fintechs are confronting increasingly sophisticated fraud across areas such as mobile wallets, cross-border commerce and rapid digital onboarding. However, our approach to fraud risk management was not built as a standalone response to a single market or to a single type of fraud.  

i2c’s PowerOne banking and payments platform brings together issuer processing, core banking, money movement and value-added services, including fraud risk management, disputes and chargeback resolutions, and reporting and analytics, across the customer lifecycle. That architecture becomes particularly valuable as the fraud environment grows more complex. Traditional approaches that depend heavily on static rules or periodically refreshed models can struggle to keep pace as fraud behaviors evolve. Siloed processes and disparate legacy technology platforms slow growth and integration, delay product launches, and expose risks to security and compliance controls.  

For financial institutions and fitnechs in Latin America, the value of that architecture lies in its ability to respond to a changing fraud environment without treating fraud prevention as an isolated function. i2c provides the context and fraud intelligence needed to inform the authorization decision in real time. 

Q: Digital payments are expanding rapidly across Latin America, while fraud schemes are becoming more sophisticated. How is i2c using AI to detect and respond to fraud in real time, and what measurable improvements are financial institutions seeing in fraud prevention and authorization performance?

A: i2c embeds fraud risk management directly within its unified banking and payments platform, enabling these capabilities to operate natively across credit, debit and prepaid issuer processing, core banking, and money movement. Our AI-informed fraud models are designed for both supervised and unsupervised machine learning detection, with an AI-informed fraud risk score delivered in real time during authorization to inform transaction decisioning. That is supported by active 24/7 monitoring of fraud alerts, anomalies, and suspicious transactions. That integration gives the model a broader context. Decisioning can draw on transactional and behavioral information such as transaction velocity, device signals, channel signals, and anomalies in spending patterns.  

This is where real-time decisioning becomes particularly important. Static, rules-based fraud systems are inherently reactive — they can only detect patterns they have already seen. Rather than relying on delayed, periodic model updates, i2c's approach incorporates dispute data as it is filed and refreshes models every three to four months, helping institutions keep pace with continuously evolving fraud behavior. AI-informed decisioning provides another way to identify patterns and anomalies as they emerge while still allowing institutions to configure fraud rules and parameters around their individual risk requirements. i2c currently supports more than 250 fraud rules in production. 

The other half of the equation is authorization performance. A fraud system that blocks fraud by declining more transactions shifts the cost onto legitimate customers rather than solving the real problem. The objective is precision: catching more fraud while approving more of the transactions that should go through. The results demonstrate that balance: i2c's fraud risk management capabilities have achieved a 40% fraud capture rate while holding customer friction to 0.5%, and authorization approval rates have reached up to 90% — exceeding industry benchmarks.  

Q: Latin America has highly diverse payment markets, regulatory frameworks and consumer behaviors. How does i2c adapt its fraud-management technology to the different requirements of markets such as Brazil, Mexico, Colombia and other countries in the region?

A: We support diverse requirements across Latin American markets through highly configurable technology that can be adapted to an institution’s specific program and risk requirements. That flexibility matters because regulatory environments, customer behaviors, product mixes and risk profiles vary across the region, so institutions need the ability to configure fraud controls around their own requirements. 

i2c’s technology was built using reusable, pre-coded building blocks rather than a hard-coded development approach. More than 100,000 pre-coded building blocks or payment functions can be accessed through APIs or self-service interfaces, allowing clients to configure capabilities without having to wait for new development. Within fraud specifically, institutions can configure rules at the portfolio, customer or account level, using either pre-configured templates or self-configurable parameters. 

That configurable technology is layered onto i2c’s unified platform, which provides the underlying foundation across what we do. The fraud intelligence itself continues to adapt as patterns change. Our fraud models draw on transaction data alongside customer behavior, merchant risk ratings and acquirer profiles, while adaptive learning mechanisms adjust thresholds and fraud rules based on recent fraudulent activity. Models are also refreshed every three to four months, helping institutions respond as fraud patterns evolve. 

Q: AI-based fraud detection can sometimes create a trade-off between stronger security and legitimate transactions being declined. How does i2c use AI to improve fraud detection while minimizing false positives and preserving the customer payment experience?

A: There is always a balance to strike between fraud and friction. If fraud controls are too permissive, institutions and their customers face greater exposure to losses. If they are too restrictive, legitimate customers absorb the cost through false positives, unnecessary authentication or declined transactions. The objective is precision: identifying fraudulent activity while allowing legitimate transactions to move through with as little friction as possible. 

For i2c, that precision starts at authorization. Our AI-driven fraud risk management solution is embedded directly within our unified banking and payments platform and evaluates risk in real time as a payment is initiated, rather than after funds have already been approved. The solution analyzes transaction data alongside customer behavior, merchant risk ratings and acquirer profiles for every transaction, giving institutions broader context to distinguish legitimate activity from behavior that warrants intervention. 

Maintaining that precision also means continuously expanding and updating the context behind each decision. The solution incorporates token provisioning data from major digital wallet providers and brings dispute data into the process as soon as it is filed, rather than relying on delayed annual updates. Models are refreshed every three to four months, helping institutions respond as fraud patterns change. 

The next piece is translating that intelligence into decisions quickly and transparently. i2c uses gradient boosted trees to support low latency, accuracy and explainability, while adaptive learning mechanisms automatically adjust thresholds and fraud rules based on recent fraudulent activity. Institutions can also configure controls around their own risk requirements rather than applying the same approach to every transaction. 

Technology alone, however, is not enough to maintain that level of precision as fraud evolves. Our data science and fraud operations teams operate in a continuous feedback loop, supported by more than 50 in-house data scientists positioned between operations and development. That feedback is paired with a broader view of the payments ecosystem. Rather than looking only at cardholder activity, the approach assesses customers, merchants and acquirers together to provide a more holistic view of risk. 

Ultimately, the measure is whether all of that intelligence translates into stronger fraud detection without introducing unnecessary friction into the customer experience. i2c's platform has demonstrated the ability to capture up to 40% of fraud volume while maintaining approximately 0.5% customer friction, contributing to reductions in fraud-related operational costs of up to 40%. 

Q: Beyond fraud prevention, where do you see the biggest opportunities for AI to transform banking and payments in Latin America? Could the same infrastructure eventually support areas such as credit decisioning, dispute management, personalization or financial inclusion?

A: Fraud prevention is an important application of AI, but the bigger opportunity lies in using that same intelligence across the entire customer relationship. We have watched AI evolve from rule-based systems to machine learning that detects patterns to agentic AI, systems that can reason, act autonomously, orchestrate across systems and learn from outcomes. Generative AI gives us reasoning; agentic AI adds perception, action and memory. 

The opportunity comes from giving that intelligence a complete view of the customer. AI gets more useful the more an institution can see across products and interactions, rather than looking at each account or transaction type in isolation. That requires an architecture where data and systems are connected. Our fraud risk management systems, for example, can read signals across a customer’s relationship and continuously learn from outcomes. That same data loop can help institutions identify which portfolio segments are worth proactively engaging, personalize interactions and make more informed decisions at the moment a customer needs them. 

Credit is a good example of where this can go next. Our decision engine already draws on on-us and off-us data with dynamic spending controls, and the same real-time infrastructure that supports fraud decisioning creates room for more responsive underwriting. The opportunity is to bring together a broader set of signals, make the decision closer to the customer interaction and continuously improve those decisions as the system learns. 

Disputes are another example. Chargeback filing, provisional credit and case adjudication are already automated end-to-end today. We are seeing a similar evolution in fraud risk management, where an anomaly can be detected, action initiated and the customer engaged within the transaction flow. That is more than automating an individual step. Increasingly, AI can help reimagine the workflow itself. 

So the common thread is not applying a single technology, like AI, to one more process. It is using connected data and real-time decisioning to act on insights within the customer experience. That creates opportunities across underwriting, servicing, personalization, payments and other areas where institutions need to make increasingly complex decisions at scale. 

Q: Looking ahead, what are i2c’s priorities for Latin America over the next 12–24 months? Are you primarily focused on expanding your existing issuer-processing and fraud-management relationships, or do you see opportunities to launch new products and capabilities specifically designed for the region?

A: Our priority is to give financial institutions and fintechs the flexibility to evolve as quickly as their markets do. That is particularly important in Latin America, where digital adoption, changing customer expectations and increasingly sophisticated fraud are creating both opportunities and new operational demands for banks and fintechs. 

i2c already has a meaningful footprint across the region. Our global network includes markets such as Mexico, Belize, Peru, Brazil, the Dominican Republic and other Latin American and Caribbean markets, backed by four global operating and service centers and two redundant active data centers that support 24/7/365 coverage. The infrastructure behind that footprint is built with headroom: i2c's platform is horizontally scalable and currently runs at roughly 40% of capacity — the platform is designed so that institutions' growth doesn't require them to repeatedly replace or bolt on new technology as their needs change. 

That's really the story here: not what's next on i2c's roadmap, but how the platform lets institutions in the region grow on their own terms. It creates room for institutions to deepen existing programs while extending into new products, propositions, segments and markets. i2c's implementation model is explicitly designed to establish foundational services that can support future value, followed by the ability to build and deliver new propositions as institutions expand. 

For Latin American institutions, the value is that they do not necessarily have to choose today between a processing relationship and tomorrow's next capability. The underlying platform is designed to allow programs to evolve as the institution's strategy, customers and market evolve.