Falcon launches MerlinFX AI agents to automate banking transactions in real-time

AI is moving deeper into banking operations, with infrastructure company Falcon launching seven AI agents that can act directly within a bank’s real-time transaction processing system rather than simply analysing data exported from the core.

Called MerlinFX, the suite is designed to automate tasks ranging from reconciliation and fraud monitoring to customer service, dispute handling and card inventory planning. Two of the seven agents—ReconFX and SurgeFX—are already in production on Falcon’s CardFX and CreditFX platforms at banks including Karur Vysya Bank and Jana Small Finance Bank, while the remaining five are being rolled out to existing clients.

The move comes as banks handle rapidly rising transaction volumes across cards and digital payments. UPI processed 24,162 crore transactions worth Rs 314 lakh crore in FY26, while outstanding credit cards rose 9.9% year-on-year to 122.86 million in July 2026, according to RBI data.


Falcon said MerlinFX is different from AI applications that sit outside the core banking environment and analyse information after it has been exported. Its agents operate within Falcon’s real-time processing infrastructure, where cards, loans, wallets and UPI transactions run on a common event stream.

That allows the agents to respond as transactions happen, rather than waiting for data to be processed later.

“Banks do not have an intelligence problem. They have a latency problem. Their underlying systems were built for end-of-day processing, and adding AI on top does not change when the data arrives,” said Priyanka Kanwar, co-founder and CEO of Falcon.

“MerlinFX puts agents to work as transactions happen, while ensuring that a bank officer approves every consequential action,” she said.

The seven agents cover different parts of bank operations. SurgeFX identifies customer segments, assesses the economics of campaigns and launches offers, while ReconFX matches transactions across networks, switches and ledgers and prioritises discrepancies based on their value.

RiskFX monitors repayment, utilisation and transaction behaviour to identify potential delinquency, non-performing assets and fraud. DisputeFX classifies chargebacks, UPI reversals and failed transactions, prepares evidence and drafts resolutions.

ResolveFX handles routine customer requests through WhatsApp, SMS or voice and escalates exceptions. ConciergeFX manages services for premium cardholders, including expiring rewards, while OpsFX forecasts branch-level card demand and reallocates inventory before new orders are placed.

Falcon provides infrastructure for credit cards, Credit Line on UPI, prepaid and UPI programmes for banks and non-banking financial companies. Its platform processes more than 15 million cards and accounts.

The company said the agents operate on the bank’s ledger and follow rules and limits set by each institution. Actions involving money movement, credit limits, account status or customer commitments are referred to a designated bank officer under the bank’s maker-checker process.

Existing Falcon clients can activate the agents through configuration, while institutions using other core systems can connect through APIs.

Security and control remain a key part of the deployment. MerlinFX operates within Falcon’s framework, certified to PCI DSS, PCI 3DS, PCI PIN and ISO/IEC 27001 standards, and undergoes annual CERT-In-empanelled testing. It can be deployed on cloud or on-premises infrastructure.

Falcon said AI inference takes place in India within the bank’s environment and client data is neither pooled nor used to train shared models. Recommendations and approvals are recorded in an immutable audit trail, while banks retain control over approval routes, budgets, thresholds and risk limits.

The larger shift is from AI being used as an analytical layer around banking systems to AI agents being embedded into the transaction workflow itself—automating routine decisions while leaving consequential actions under human control.