IBM Says Tokenization Key to Enabling Autonomous AI-Driven Finance

Tokenization’s next big customer may not be human. It may be an AI agent—and that shift could force the financial industry to rebuild its rails around 'machine-executable' money, according to a new report from IBM’s Institute for Business Value (IBV).

In its study on banking in the tokenized economy, IBM argues that 'agentic AI'—autonomous systems capable of interpreting goals, making decisions, and executing tasks—should not be treated as a separate technology wave from tokenization. Instead, the two are increasingly complementary: as AI becomes capable of running financial workflows end-to-end, tokenization becomes the infrastructure that makes assets, settlement, and compliance readable and actionable by machines.

The premise is straightforward. AI can automate decisions, but it cannot autonomously operate in a world built for human paperwork, manual approvals, limited banking hours, and layered intermediaries. Tokenization changes that by encoding ownership, rights, conditions, and settlement rules into digital form—often paired with smart contracts that can execute transactions and settle them automatically. In IBM’s framing, once AI becomes the “operator” of finance, it will demand a new set of 'programmable rails' to move value at scale. Tokenization is that rail.

From chatbots to financial operators

IBM’s report positions agentic AI as more than customer-service chatbots or analytics tools. In finance, it can expand into underwriting, risk assessment, payments, settlement, treasury operations, portfolio rebalancing, supply-chain finance, and corporate cash management—tasks traditionally constrained by internal controls and fragmented infrastructure.

But full autonomy requires more than compute. It requires financial systems designed for automated interpretation and execution. Tokenized assets can provide standardized, machine-readable representations of value, while distributed ledgers can improve traceability. Smart contracts can embed business logic directly into transactions, allowing AI systems to verify criteria and trigger settlement without manual coordination.

Four intersections: incentives, efficiency, security, and agent-to-agent commerce

IBM maps the convergence of tokenization and agentic AI across four key areas.

First is 'incentive alignment' through monetization and performance-based rewards. If the outputs, interactions, or services of AI agents are represented and compensated in tokens, AI systems can participate directly in network economies—earning rewards for tasks, coordinating with other agents, and distributing value programmatically. The implication is a shift from platform economies powered by people to ecosystems where machines also become economic actors.

Second is efficiency and compliance. Settlement, verification, reporting, and regulatory checks are among the most time-consuming parts of financial operations. Tokenized assets and smart contracts can automate these processes, enabling AI agents to validate transaction conditions, assess risks, check regulatory requirements, and execute trades in a single workflow. IBM emphasizes that automation alone is not enough; 'auditability' matters. When transactions are recorded on a ledger and governed by predefined rules, actions taken by AI agents can be reviewed, tested, and investigated—an important requirement for financial institutions grappling with accountability and control.

Third is security. Autonomous agents moving funds expands operational velocity but also raises the stakes for trust. Tokenization, combined with cryptographic protections and immutable records, can reduce tampering and fraud risks. IBM suggests tokenized systems can embed security at the settlement layer—providing guardrails for AI-driven activity.

Fourth is agent-to-agent commerce. IBM expects an increasing share of machine-driven transactions—agents buying and selling data, requesting services, allocating resources, and completing payment without human intervention. In that scenario, tokenization becomes more than a financial innovation; it becomes a transactional language for an emerging autonomous economy.

Executives see tokenized settlement as an autonomy multiplier

Survey results cited in the report reinforce the thesis that settlement infrastructure is becoming the bottleneck for AI-led finance. IBM said 57% of surveyed executives believe tokenizing settlement rails would significantly strengthen the autonomy of agentic AI. Another 61% pointed to clear synergies around cross-platform interoperability between agentic AI and tokenization, while roughly 60% expect AI to materially influence transparency, auditability, and programmability in settlement networks.

Interoperability is central because AI agents must traverse multiple environments—payment networks, asset platforms, databases, and compliance systems—to execute financial tasks. Closed architectures limit AI mobility. Tokenized assets and programmable settlement networks, by contrast, can give agents a consistent way to recognize assets and transact across systems.

The report highlights 'machine-readable assets' as a key design goal—structures that AI can interpret and execute without relying on human-facing contracts or siloed account records. In the asset and wealth management segment, 77% of respondents identified machine-readable assets as the most important value proposition of agentic AI. If assets are tokenized and standardized, an AI system could autonomously rebalance portfolios, execute conditional trades, and in some cases perform peer-to-peer exchanges without routing through legacy fiat settlement processes.

Wallets and custody move to the center of AI-native finance

If AI agents are to transact, they need wallets—but not the kind designed for consumer apps. IBM frames the AI wallet as an execution layer tying together identity, permissions, asset positions, transaction history, and payment functionality. The practical questions quickly become institutional: which agents are authorized to do what, within what limits, under which approval conditions, and with what accountability if something goes wrong.

That elevates digital custody from a back-office service to a core control point. As AI agents handle tokenized assets, key management and access control become systemically important risks. Banks and custodians could become gatekeepers for AI-led financial activity, leveraging traditional strengths such as compliance, know-your-customer processes, anti-money-laundering controls, and institutional asset safeguards.

At the same time, IBM suggests the competitive landscape could shift. If incumbent institutions fail to establish wallet and custody infrastructure, customer relationships may migrate toward fintech firms and blockchain-native platforms. In an AI-automated future, controlling accounts may matter less than controlling wallets, permissions, and the operational rules that govern value movement.

Supply chains, robotics, and the new role of payments

IBM’s most concrete near-term scenario appears in corporate finance. As companies redesign workflows around AI automation, AI agents could interact directly with supply chains via tokenized infrastructure: automatically placing orders when inventory drops, triggering invoicing when delivery conditions are satisfied, and settling payments in real time through smart contracts.

This is not simply process automation. It implies a re-architecture of enterprise operations where procurement, logistics, accounting, payments, and financing converge into a single programmable flow. In such a world, supply-chain finance shifts from after-the-fact reconciliation to real-time execution—payments becoming not just a financial function but a core component of industrial infrastructure.

IBM characterizes this as financial services enabling robotics economically—an underpinning for the next phase of industrial transformation. If autonomous robots and AI agents can buy resources, request services, and pay costs on their own, the question becomes strategic: who provides the settlement rails for an autonomous economy?

Risk expands alongside autonomy, with quantum threats on the horizon

The report also warns that combining agentic AI with tokenized finance increases the attack surface. Once agents hold wallets, execute smart contracts, and move assets automatically, risks range from credential compromise and malicious smart contracts to data manipulation, automated market abuse, and high-speed fraud.

IBM also flags quantum computing as a longer-term systemic risk. It said 89% of surveyed executives believe quantum threats are already amplifying system risk. Because blockchain and digital asset security relies heavily on cryptography, quantum-capable attacks could undermine widely used security assumptions—raising the urgency for financial institutions to begin planning for quantum-resistant approaches.

Regulators face a parallel challenge: if AI agents—not humans—execute trades and transfers, accountability frameworks may need to be redesigned. Standards for market abuse prevention, internal controls, and auditing become more complex, particularly in decentralized environments where enforcement and oversight are harder to apply consistently.

Tokenization as a prerequisite, not a product feature

IBM’s core message is that tokenization should not be viewed narrowly as an upgrade for stablecoins or real-world asset (RWA) digitization. It is, in its view, a prerequisite layer for AI systems to operate inside financial markets. If AI is expected to decide and execute, then assets and settlement networks must be legible to AI and capable of automated execution.

Just as importantly, IBM suggests agentic AI could become tokenization’s strongest demand driver. Machines operate 24/7, process real-time data, and execute instantly when conditions are met—behavior that clashes with slow, batch-based reconciliation systems. Programmable tokenized rails offer a better fit for that tempo.

For banks, the implication is competitive and structural: building token-ready core systems, digital custody, wallet infrastructure, and smart-contract operational capability is not merely a play for stablecoin adoption. It is positioning for a broader contest over who will operate the settlement layer of an emerging autonomous economy.

In IBM’s framing, if AI is finance’s brain, tokenization becomes its hands and feet. As the two converge, finance could move from a model where humans initiate and banks process, to one where machines decide and networks settle—marking what the report describes as the opening act of autonomous finance.