Skip to Content

Franken-core: AI Agents and Programmable Money Force Core Banking Overhaul

Dharmesh Mistry warns legacy systems cannot absorb billions of agentic micro-transactions
Sk Jabedul Haque
Jun 27, 2026 5 min read 7 views
Franken-core: AI Agents and Programmable Money Force Core Banking Overhaul
Navigation
10 Sections
    Banks face a Franken-core crisis: legacy systems patched with fintech cannot handle billions of AI agent micro-transactions. Dharmesh Mistry warns the endgame is not a better core but a different kind of bank where programmable money and agentic AI replace traditional ledgers.

    Franken-core has become the defining metaphor for core banking technology in 2026. Dharmesh Mistry, a thirty-year banking veteran and FinTech Futures columnist, published a three-part series this week arguing that the patchwork of legacy cores and bolt-on fintech layers cannot absorb the transaction volume coming from AI agents operating on programmable money rails.

    What Happened

    On June 26, FinTech Futures published the latest instalment of Mistry’s Franken-core series, declaring that “the AI agents are coming, the transaction volumes are coming, and the programmable money infrastructure is being built right now, with or without you.” The article cites the University of Cambridge’s 2026 Global AI in Financial Services Report: 81 percent of respondents expect agentic AI to be mainstream by 2030, while 52 percent have already integrated the technology. BNP Paribas Polska’s IT leader Dariusz Flisiak confirmed the trend: “For me, the biggest trend I’ve seen this year is agentic AI. Agents read documents, research and analyse databases, create hypotheses, verify those hypotheses, and only then come back with an answer that is well thought and fact checked.”

    The scale challenge is concrete. AI agents making billions of micro-transactions daily on stablecoin rails—programmable money that settles in real time and carries conditional logic natively—exceed what any Franken-core or second-core architecture can process. Mistry argues the endgame “isn’t a better core; it’s a different kind of bank. One where the ledger thinks, the money flows, and the AI agents operate autonomously.”

    Why It Matters

    The implications stretch beyond banking technology vendors. When AI agents can execute payments autonomously, traditional compliance, settlement, and reconciliation workflows collapse. Google Cloud’s Agent Payments Protocol (AP2) combines programmable payments via modern blockchains like Sui with open protocols such as A2A and MCP. Wipro’s programmable money platform uses blockchain and agentic AI to automate financial operations and enable autonomous transactions. Circle’s research emphasizes that banks must determine whether an agent is allowed to act, under what limits, and with what governance—a fundamental shift from verifying human intent to governing agent autonomy.

    First Commerce Bank in New Jersey is already migrating to FIS Horizon specifically to accelerate AI adoption, with the partnership highlighting “standardised data feeds and agentic commerce tools.” Backbase’s acquisition of Kasisto aims to deploy agents that natively handle the full arc from customer intent to governed resolution across chat, messaging, and voice. Oracle’s banking roadmap calls for lightweight, composable cores that decouple transaction processing so banks can rapidly plug in sophisticated AI agents.

    What’s Next

    The transition from Franken-core to intelligent ledger will not be a single migration. Banks face a choice between incremental modernization—wrapping legacy systems with API layers—and full transformation to cloud-native, AI-native architectures. The Bank for International Settlements has already examined how generative AI can assist in managing cash and liquidity in real-time gross settlement systems. Stablecoin infrastructure becomes the backbone of this new financial architecture, as central banks themselves are preparing for agentic finance. As the GENIUS Act establishes the U.S. regulatory framework for stablecoins, programmable dollars become the default rail for agentic commerce.

    For technology leaders, the priority shifts from core replacement to building the programmable money layer that sits above it—stablecoin issuance infrastructure, agent identity and governance frameworks, and composable settlement rails. The monster in the basement is not dying quietly, but the blueprint for what replaces it is already in plain sight.

    Related coverage: Big Tech AI Demand: 50B Spending Spree Meets Customer Reality Check, Big Tech AI Spending: The $2.7 Trillion Bill Comes Due, Bittensor Unveils 18-Month Decentralization Roadmap, BitGo Cuts 15% Workforce: Crypto Custodian Pivots to Stablecoins and AI, On-Chain Convergence: Regulated Rails Redefine Global Finance, Range Raises $8.3M Series A: Stablecoin Infrastructure Funding, MoonPay Acquires Entendre: AI Accounting Agents for Stablecoin Finance.

    Frequently Asked Questions

    Franken-core is Dharmesh Mistry's term for the patchwork of legacy core banking systems held together by fintech API layers. Most banks run on decades-old mainframes wrapped with modern interfaces, creating a fragile architecture that cannot scale to handle AI agent transaction volumes.
    AI agents on programmable money rails can execute billions of micro-transactions daily. Neither the legacy Franken-core nor the second-core middleware layer can process this volume at real-time speeds with conditional logic.
    Programmable money refers to stablecoins and bank tokens that carry conditional logic natively—payroll code that automatically executes transactions within specific parameters. Examples include USDC on Ethereum, Sui, and other blockchains where settlement is instant and programmable.
    First Commerce Bank in New Jersey is migrating to FIS Horizon for AI capabilities including agentic commerce tools. Backbase acquired Kasisto to deploy agents across conversational surfaces. Oracle and Wipro are building composable cores and programmable money platforms for agentic finance.
    Mistry argues the endgame is not a better core but a different kind of bank: one where the ledger thinks, money flows autonomously, and AI agents operate on programmable settlement rails. The programmable money layer—stablecoin infrastructure, agent governance, composable settlement—sits above the core.
    Sk Jabedul Haque

    Sk Jabedul Haque

    Founder & Chief Editor

    Building India's most trusted finance education platform — simplifying news, calculators, and market trends so anyone can understand and invest confidently.