Agentic Payments Arrive in Production: What Blik’s First AI-Agent Transaction and the New Know-Your-Agent Framework Mean for Finance Teams

Last Updated: September 11, 2026
By the Kurums.com Finance Desk

What Just Happened With Agentic Payments?

Blik, Poland’s dominant mobile payment network, confirmed it has processed the first agentic payment transaction on its rails: a user pre-authorized an AI agent to search for a specific product and complete payment automatically the moment it became available.

This is a meaningful shift from earlier “AI shopping assistant” demos, most of which stopped at generating a product shortlist or a draft cart that a human still had to check out manually. In Blik’s case, the agent held standing authorization to both search and pay within pre-set conditions, without a human confirming the final transaction. For finance teams, that single detail changes the risk conversation from “AI helps a customer shop” to “an autonomous system is initiating an actual movement of money,” which is a fundamentally different category of operational, fraud, and reconciliation risk.

What Is the Know-Your-Agent (KYA) Framework?

Know-Your-Agent, or KYA, is a proposed interoperability standard from Ant International, Visa, and Mastercard designed to identify, verify, and onboard AI agents consistently across card networks, digital wallets, and marketplaces.

KYA is explicitly modeled on Know-Your-Customer (KYC) obligations that banks already run for human account holders, extended to a non-human actor that can now initiate a payment. The framework aims to answer three questions before an agent is allowed to transact: which platform or company deployed this agent, what spending limits and merchant categories is it authorized for, and how can a receiving bank or merchant confirm the agent’s authorization has not been revoked or spoofed. Three of the largest players in global payments backing the same framework at the same time is a strong signal that card networks expect agent-initiated transactions to scale quickly enough that fragmented, network-by-network agent verification would become unworkable within a year or two.

Why Is Consumer Trust Still So Low?

New Visa research found that only 23% of US consumers trust generative AI to handle payment transactions on their behalf, even as AI-assisted shopping and product discovery adoption keeps climbing.

That gap between behavior and trust is the central tension finance and product teams are navigating right now. Consumers are comfortable letting an AI agent research, compare, and recommend, but handing over the actual payment step is a different psychological threshold — closer to handing someone a credit card than asking them for advice. Two things are likely driving the gap: consumers have limited visibility into what an agent is authorized to spend and on what, and there is no widely understood, bank-grade recourse process yet for an agent-initiated transaction gone wrong, comparable to a chargeback. Until frameworks like KYA translate into consumer-visible controls — spending caps a person can see and adjust, real-time transaction alerts, and a clear dispute path — trust is likely to keep lagging usage.

Why Is Funding Flowing Into Agentic GRC Platforms Right Now?

Agentic governance-risk-compliance platform HelmGuard raised $7.3 million in seed funding this week, using AI agents to automate governance, risk, and compliance monitoring inside regulated sectors including financial services.

The timing is not a coincidence. As agent-initiated payments move into production, compliance teams need tooling that can monitor agent behavior at the same speed the agents themselves operate — a rules engine or a quarterly audit cycle cannot catch a misbehaving agent that can execute thousands of transactions in an hour. Investors backing agentic GRC platforms are effectively betting that “AI to watch the AI” becomes a standard line item in compliance budgets over the next 12 to 24 months, in the same way fraud-detection software became a standard line item once card-not-present transactions scaled in the 2000s.

What Does This Mean for Finance and Treasury Teams?

Finance and treasury teams should treat agent-initiated payments as a new transaction category requiring its own authorization limits, monitoring, and reconciliation logic, rather than folding it into existing card-not-present or automated-payment policies.

Four practical steps are worth prioritizing now, ahead of broader adoption. First, define an explicit spending envelope for any AI agent a company authorizes to transact on its behalf — per-transaction cap, daily cap, and an approved merchant category list — the same discipline already applied to corporate card programs. Second, confirm with payment processors and card networks what agent-verification standard, such as KYA once finalized, they plan to support, since a business accepting agent-initiated payments will eventually need to distinguish a verified agent transaction from an unverified one at the point of sale. Third, build agent-specific fraud monitoring rather than assuming existing fraud models transfer cleanly, since an agent’s “normal” transaction pattern — rapid, narrow-purpose, time-triggered — looks anomalous under models trained on human purchasing behavior. Fourth, revisit dispute and chargeback procedures with the assumption that a customer may need to contest a transaction their own authorized agent initiated, which is a scenario most current dispute workflows were not built to handle.

What Changes for Accounting and Reconciliation?

Agent-initiated purchases complicate reconciliation because the transaction record no longer maps cleanly to a single human decision-maker, which is the assumption most expense and accounts-payable workflows are built on.

Finance teams running procurement or expense software should expect a near-term need to capture an additional data field alongside every transaction: which agent, under whose authorization, executed the purchase, and against which pre-approved policy. Without that field, an auditor reviewing a spend report a year from now has no way to distinguish a legitimate agent-initiated purchase from an anomaly, and internal controls built around “who approved this” break down when the answer is “an agent, acting on a standing policy” rather than a named employee. Finance software vendors serving mid-market and enterprise clients are the ones most likely to ship this capability first, since their customers face the compliance pressure earliest.

Key Takeaways on Agentic Payments

Is agentic payment technology now live for ordinary consumers? Partially — Blik’s transaction shows the capability is technically live on at least one major payment network, but broad consumer rollout is still gated by trust, regulatory clarity, and merchant readiness.

Does KYA replace existing KYC obligations for banks? No — KYA is designed to sit alongside KYC, adding a verification layer specifically for non-human agents acting on behalf of an already-verified human or business customer.

Are agentic payments currently regulated the same way as card payments? Not specifically — most jurisdictions are still applying existing card and payment-services rules to agent transactions rather than having agent-specific regulation in force.

Should a mid-size business start accepting agent-initiated payments now? Most payment and risk advisors recommend waiting for a published verification standard such as KYA to stabilize before actively marketing agent-payment acceptance, while still monitoring the space closely.

How Does This Fit Into the Broader Fintech Picture?

Agent-initiated payments are arriving alongside several other structural shifts in how money moves in 2026, and finance teams evaluating one typically need to evaluate all of them together.

Kurums.com’s guide to Embedded Finance Trends 2026 covers the broader restructuring of where and how payment functionality gets built into non-financial products, which is the same infrastructure layer agentic payments plug into. Businesses evaluating whether to hold reserves or settle in stablecoins alongside agent-payment rails should also see Stablecoins and Business Finance in 2026 for why adoption still trails infrastructure readiness — a dynamic that closely mirrors the consumer-trust gap in agentic payments. For the full range of banking, payments, and fintech regulation coverage, visit the Kurums.com Finance department hub.

Frequently Asked Questions About Agentic Payments

What is an agentic payment, exactly?
An agentic payment is a transaction initiated and completed by an AI agent acting under a standing authorization from a human or business, without a person approving that specific transaction at the moment it occurs.

How is this different from a recurring subscription payment?
A recurring payment executes a fixed amount on a fixed schedule that a person set in advance, while an agentic payment involves the agent making a real-time decision — what to buy, when, and sometimes from which merchant — within broader limits a person defined.

Who is liable if an AI agent makes an unauthorized or mistaken purchase?
Liability frameworks are still being worked out; card networks developing standards like KYA are expected to define clearer liability allocation between the agent platform, the card issuer, and the merchant over the next one to two years.

Can businesses block agentic payments from being used at checkout?
Yes, in most current implementations a merchant or payment processor can choose not to accept agent-initiated transactions, similar to how some merchants restrict certain card types today.

Is agentic GRC software only relevant to large financial institutions?
No — any business authorizing AI agents to transact on its behalf, including mid-size e-commerce and B2B companies, faces the same monitoring need, though enterprise financial institutions are adopting agentic GRC tooling fastest due to regulatory exposure.

Sources

• Finextra — Blik records first agentic payment transaction
• Finextra — Ant International, Visa and Mastercard to develop ‘Know-Your-Agent’ framework
• Finextra — Consumer trust in agentic payments continues to lag
• Finextra — HelmGuard raises $7.3m for agentic GRC platform