Evaluating identity verification partners? 6 questions for payments leaders at Money20/20

Money20/20 USA 2026 arrives as several forces converge on the payments industry. Agentic AI is initiating transactions without humans. Cross-border rails are multiplying. Embedded finance has scattered the customer relationship across dozens of partners, and real-time payment rails have compressed decision windows to milliseconds.

With one in three attendees holding C-suite titles, vendor conversations won’t be won on feature lists or fraud detection rates alone. Payments leaders should push vendors on the numbers that move the P&L: conversion rate, fraud that slips through (FAR), genuine customers wrongly turned away (FRR), and verification latency.

We sat down with Raul Liive, to turn his latest guidance and recent insights on AI buying agents from Veriff’s leadership into a direct Q&A on how to evaluate identity and fraud vendors this year.

1. Why is agentic commerce the biggest unresolved identity problem for payments leaders?

Agent protocols can verify software, but they don’t prove who is behind it. When an agent makes a payment, vendors need to establish that the human is real, the agent is authorized, and the specific transaction is within that authorization.

You have to ask vendors how they bind agent activity to a verified person, not just an authenticated session. A major green flag is a vendor with a purpose-built model for agentic transactions that can enforce action-level scope boundaries and produce a queryable audit trail linking the human, the agent, and the authorized action. The red flag is a vendor claiming that standard API signatures or existing human verification flows already solve this.

2. Can defenses keep up with generative AI without hurting pass rates?

Yes, but only if you actually measure the tradeoff. Fraudsters aren’t perfecting fake passports anymore; they’re taking over real accounts and letting automation handle the volume. If your model is tuned too aggressively to block automation, you’ll block legitimate agents acting on behalf of your best customers.

As our CTO Hubert Behaghel recently put it, it is a ten-times bigger business concern to reject a good person than to let a fraudster through. A wrongly blocked agent is just an abandoned basket that routes your customer straight to a competitor.

When evaluating a vendor here, the green flag is when they measure agent traffic separately, tracking approvals and false rejections apart from human traffic, and can show you the false positive rate on real users. If they quote a 99% deepfake catch rate but conveniently leave out how many good users they block, that’s a massive red flag. They are only showing you one column of the P&L.

3. How do we maintain trust after onboarding?

 Onboarding is where identity trust is either established or compromised for the life of the account. Because of synthetic identities and account takeovers, passing a single document check isn’t enough to prove someone is a real, unique person months later.

The mistake organizations make is treating onboarding as a finish line rather than a baseline. A green flag vendor makes verification re-runnable. They combine document authenticity, liveness, and behavioral signals at onboarding, and then use targeted biometric re-authentication at moments of consequence, costing a real customer seconds, but destroying a fraud ring’s economics. The red flag is a vendor relying on a single day-one check or touting improved fraud detection without disclosing how much legitimate completion time and pass rates suffer.

Meet Veriff at Money 20/20

Stop fraud in seconds: See Veriff live at Money 20/20

Heading to Money 20/20? Meet our team on the show floor to learn how Veriff balances rock-solid security with a frictionless onboarding experience.

Book a meeeting with Veriff

4. How should platforms handle rapidly changing regulations?

KYC and AML requirements vary by jurisdiction and change constantly. The cost of a slow adaptation isn’t just regulatory risk; it’s operational downtime. Every week a workflow sits out of compliance in a new market is a week of delayed revenue and expensive manual workarounds.

You need to ask vendors how long it actually takes to adjust a workflow when a rule changes. The green flag is a configurable compliance workflow that your internal team can adjust for new markets without waiting on the vendor’s engineering cycle. If every regulatory change requires a custom engineering ticket on their end with no guaranteed SLA, that is a glaring red flag.

5. How are real-time payment rails changing identity verification?

Real-time rails compress the decision window to milliseconds. A verification process that adds even a few seconds of friction is the difference between a completed transaction and an abandoned one. And once the money moves, it rarely comes back.

This requires balancing fraud loss, blocked legitimate transactions, and operational cost. Slower verification drives up manual review latency and suppresses your pass rate right when speed is critical. Look for the green flag: a vendor that shares production speed benchmarks specific to real-time rails alongside false positive rates at that speed. Generic speed claims based on standard checkout flows are a red flag.

6. Does embedded finance create new identity gaps?

Embedded finance spreads the customer relationship across multiple partners – the platform, the BaaS layer, the payment provider. Each handoff is a potential gap where verification can be duplicated, skipped, or handled inconsistently.

The liability exposure here is massive because no single partner naturally owns accountability. You must ask vendors how they coordinate across a full stack. A green flag vendor has a defined model for this coordination, clear escalation paths, and visibility into standard enforcement across integrations. If they only address verification for their own direct integration and defer downstream accountability, that’s a red flag.

Where does Veriff stand on these questions?

 We’re very direct about our tradeoffs. We combine document authenticity, liveness, and behavioral signals, and we publish both our fraud catch rates and false positive rates so you see the full conversion impact. Our compliance workflows are configurable without waiting on our dev cycles, and verification speed on real-time rails is a core product requirement for us.

Regarding agentic commerce: we are honest. It is a genuinely new problem, and no vendor at Money20/20 this year, including us, has a fully mature answer yet. We will tell you exactly where our capabilities stand today and what we are building toward. Vendors claiming they’ve completely solved it are overstating their readiness just to close a deal.