On July 1, nsKnox launched AI Agent Caller, an autonomous voice agent that places vendor verification calls on its own. The product joins the company's PaymentKnox suite and goes after one of the most stubborn manual steps in corporate finance: the callback a payments team makes to confirm a vendor's bank details before money moves. (nsKnox announcement)
The idea is straightforward. Instead of a person picking up the phone to verify a payee, the agent schedules the call, speaks the vendor's language, runs the conversation, and writes the whole exchange to an audit log. nsKnox says the recording and summary land on the same screen as its banking-based validation checks, so an accounts payable team sees the voice confirmation and the data confirmation together.
What nsKnox actually shipped
AI Agent Caller is available now. Per the company, it handles autonomous, multilingual calls, schedules them inside the payee's local business hours, and holds context-aware conversations rather than reading a rigid script. It verifies identity without asking the vendor for full banking details, records the call, writes an automatic summary, and logs everything to the PaymentKnox Work Canvas. A user can trigger a call straight from a payee record, and the result appears next to the banking-data checks nsKnox calls Quick Check and KnoxVerify.
Chief executive Nithai Barzam framed the logic this way: "Validation based on authoritative banking data is the gold standard in payment security. It is the only source of truth that fraudsters cannot manipulate or impersonate." The voice call, in other words, sits on top of a data check, not in place of it.
Why the vendor callback is a fraud chokepoint
The callback exists because a bank-account change request is the classic Business Email Compromise move. A fraudster poses as a known vendor and asks finance to send the next payment to a new account. A quick verbal check to a trusted number is meant to catch it before the wire leaves.
The money at stake is not small. The FBI Internet Crime Complaint Center logged about 2.8 billion dollars in reported BEC losses in 2024 and about 3.04 billion dollars in 2025. Its public alert put global exposed BEC losses near 55 billion dollars between 2013 and 2023. (FBI IC3) Most of that money moves by wire or ACH, which makes it fast and often impossible to claw back. nsKnox cites 43 billion dollars in BEC losses over five years and says 83 percent of large enterprises have been targeted by payment fraud.
Interested in implementing similar AI solutions? Discover how PATech Labs can help your business leverage cutting-edge artificial intelligence.
Learn About Our ServicesWhat an autonomous voice agent changes
The callback works, but it does not scale. It is slow, inconsistent, and quietly skipped when a deadline is close. Automating it fixes the scale and consistency gap, and the audit trail is the part that matters most for regulated finance. Every verification becomes a logged, replayable record instead of a note that someone may or may not have written down. That is exactly what an auditor or a fraud investigator asks for after the fact.
It is worth being precise about what the agent is not. It is one layer. The banking-data validation is still the source of truth, and the call adds a human-facing confirmation on top. An enterprise that treats the voice step as the whole defense would be missing the point.
What this means for banks, AP teams, and contact centers
This is the shape of voice AI that actually holds up in 2026: a narrow, high-value workflow wrapped in compliance and a clean record. Vendor callback verification is a strong template because the task is bounded, the payoff is obvious, and the log matters as much as the outcome.
- For accounts payable teams, the gain is consistency. Every payee change gets the same verification, in the payee's language, at a reasonable local hour, without a person spending an afternoon on hold.
- For banks and their corporate clients, it lowers the odds that a rushed team skips the one step that would have stopped the fraud.
- For contact center leaders, it is a reminder that the best early use cases for voice agents are the calls nobody wants to make: repetitive, script-adjacent, and bound by compliance.
At PATech we build this same pattern, a focused voice workflow with verification and an audit trail underneath. The lesson from launches like this one is that the record is a feature, not an afterthought.
The catch: autonomy needs an audit trail
An autonomous agent that calls your vendors is also a new attack surface and a governance question. Enterprises are already cautious. Through 2026, many finance and healthcare teams have paused or rolled back at least one AI deployment, and buyers now expect SOC 2, ISO 27001, and similar controls before they will even evaluate a voice vendor. An agent that verifies payments has to be at least as auditable and secure as the manual process it replaces. nsKnox leans on that by putting the recording, the summary, and the validation result in one place.
The takeaway
AI Agent Caller is a small, concrete signal of where voice AI is landing this year. Not flashy customer-facing chat, but the unglamorous, high-stakes back-office calls where consistency and a clean record are worth more than personality. That is the version of AI agents that survives a security review, and it is the version that quietly saves a company from wiring money to the wrong account.