Consumers Use AI Assistants but Won't Hand Over Their Wallets: Visa Data Reveals the Agent Payment Trust Gap

Research released by payments giant Visa on September 9, 2026 shows a huge gap between consumers' enthusiasm for AI assistants and their willingness to trust them with payments: 72% of consumers have used an AI assistant, but only 23% of U.S. consumers trust generative AI agents to handle payments on their behalf. When a payment brand steps in, the picture changes markedly—61% of consumers say they trust Visa to process agent transactions.

Visa Group President Oliver Jenkyn said bluntly in the announcement: "Trust is the foundation that drives adoption of agentic commerce. As consumers deploy AI agents to shop on their behalf, they will increasingly rely on trusted payment experiences and brands." This statement pushes payment brands from mere transaction processors to the role of trust infrastructure in the agent economy.

Tiered Trust: From "Assisted Decisions" to "Authorized Spending"

The trust gap is not a vague concept. The Global Digital Shopping Index: Agentic Commerce Deep Dive (July 2026), produced by Visa in collaboration with PYMNTS Intelligence, shows clear tiers in consumer acceptance for different tasks: 56% are willing to let agents search for and compare products, 51% are willing to let them manage loyalty programs, but only 35% are willing to let agents directly complete purchase payments. From "assisted decisions" to "authorized spending," trust falls off a cliff.

This tiered trust has direct implications for agent payment adoption. The payments industry has long relied on KYC (Know Your Customer) to build trust between users and accounts, but agent payments introduce a new entity—the AI agent itself. An agent is not a legal entity and cannot sign contracts; its behavior is determined by algorithms and user authorization. Traditional KYC frameworks cannot answer questions such as "Is this agent authorized to spend this money?" or "Has it been maliciously tampered with?" This is precisely the core of the KYA (Know-Your-Agent) concept that OceanAlt focuses on: verifying agent identity, authorization intent, and transaction behavior before settlement.

Visa's data confirms this need from the consumer side. When users face a "black box" agent, trust cannot be established; but when a brand like Visa steps in, it effectively provides a "trust anchor"—users believe the brand will conduct compliance reviews, block fraud, and protect funds. The 61% trust level shows that brand endorsement can significantly reduce perceived risk.

Three Implementation Paths for Tiered Trust Mechanisms

For the industry, this means agent payments cannot simply replicate existing payment processes; instead, tiered trust mechanisms need to be built. Specific paths may include: first, payment brands or banks acting as "trust intermediaries," guaranteeing or insuring agent transactions; second, introducing mandate mechanisms at the technical level, where users can set per-transaction limits, daily cumulative caps, and recipient whitelists, with agents able to act only within preset boundaries; third, pre-settlement screening becoming standard, with real-time screening of counterparties, on-chain taint, and sanctions lists before funds are transferred.

Visa's research also implies a key point: consumer trust in agents is not all-or-nothing, but task-dependent. This means product design should allow users to gradually delegate authority by risk level—first letting agents compare prices, then manage points, and only finally authorizing payments. This gradual trust-building aligns with the RAP (Risk-Adjusted Permissions) standard advocated by OceanAlt: dynamically adjusting agent permissions based on task risk, rather than opening or prohibiting everything across the board.

The Battle for Standard-Setting Power and the Accountability Gap

From a competitive standpoint, Visa's move is intended to seize standard-setting power in agent payments. If payment brands can become the default trust layer for agent transactions, they can occupy a key node in the future machine-to-machine payment ecosystem. But the challenges are equally obvious: agent transactions may involve cross-border, multi-currency, and high-frequency small-value scenarios, and the processing capacity and compliance frameworks of traditional payment networks may not be suited to them. New protocols such as stablecoins and x402 are trying to fill this gap, but lack brand trust endorsement. Visa's entry may accelerate integration or competition between traditional payment giants and emerging crypto infrastructure.

For consumers, the real test is: when an agent makes a mistake, how is responsibility defined? Is it improper user authorization, a flawed algorithm by the agent developer, or an oversight in the payment platform's review? If Visa's trust endorsement is not accompanied by a clear accountability mechanism, the 61% trust level may be a castle in the air. The industry needs to establish legal and insurance safeguards beyond technical compliance before consumers can truly feel comfortable handing over their wallets.

---

Original source: PYMNTS · https://www.pymnts.com/news/artificial-intelligence/2026/consumers-use-ai-assistants-hesitate-hand-over-their-wallets/