TL;DR: Merchants are increasingly using agentic AI for chargeback documentation and refund review, while passkeys are moving from pilot to practice as credential stuffing and phishing continue to drive account takeover, according to Sift and cited FIDO Alliance and NIST guidance. The governance challenge is no longer authentication alone, but trust, measurement and identity resolution across the customer journey.
At a glance
What this is: Sift says merchants are pairing agentic AI with passkeys as digital commerce teams try to reduce account takeover and make trust measurable across the customer journey.
Why it matters: It matters because identity and fraud teams now have to govern authentication, device trust, and AI-assisted review as connected parts of one control model.
Context
Digital commerce teams are moving past a narrow password replacement debate. The real problem is how to prove identity, preserve conversion, and keep fraud controls effective as customer journeys span sign-up, login, transactions, refunds, and loyalty actions.
This article sits at the intersection of identity verification, fraud operations, and AI-assisted decisioning. Passkeys change the authentication layer, agentic AI changes the workflow layer, and account takeover shows why trust now has to be measured across the full customer lifecycle.
The starting position is not unusual for merchants under sustained phishing and credential stuffing pressure. What is changing is the governance question: not whether to modernise authentication, but how to align identity trust, human oversight, and operational metrics.
Key questions
Q: What breaks when merchants rely on passwords and recovery flows for customer authentication?
A: Password-based access collapses when phishing and credential stuffing scale faster than user education or manual review. Recovery paths often become the weakest point because they can bypass strong login controls. Merchants need authentication that resists replay, plus governance over fallback paths, trusted devices, and step-up checks for high-risk actions.
Q: Why do account takeover controls now affect revenue as much as security?
A: Account takeover distorts approval rates, drives false declines, increases refund abuse, and damages loyalty economics. The loss is not only the immediate fraud event but also the long-term erosion of customer value and model quality. That is why fraud, payments, and identity teams need shared metrics instead of isolated dashboards.
Q: What are the signs that passkeys are not enough on their own?
A: Passkeys are not enough when suspicious device changes, abnormal recovery activity, or repeated step-up prompts still appear around otherwise successful logins. Those signals suggest the authentication layer is stronger than the surrounding identity trust model. The gap is usually in device trust, lifecycle controls, or post-login fraud detection.
Q: How should security teams govern agentic AI in fraud detection?
A: Start by separating detection support from decision authority. Agentic AI can correlate signals quickly, but any action that changes customer treatment, case priority, or compliance status needs explicit policy, traceability, and a human escalation path. Governance should define what the system may recommend, what it may execute, and what always requires review.
Technical breakdown
Why passkeys reduce phishing risk but not identity risk
Passkeys replace shared secrets with cryptographic assertions bound to a device and relying party, which makes phishing and credential stuffing far less effective. But they do not solve all identity problems because a valid passkey still authenticates whoever controls the trusted device and account recovery path. In commerce, that means authentication strength and identity trust are related but not identical. A secure login can still lead to fraud if device binding, step-up checks, or recovery workflows are weak.
Practical implication: Treat passkeys as a stronger authentication layer, not as a complete fraud control.
How agentic AI changes fraud and operations workflows
Agentic AI is not just automation. It is software that can plan tasks, take actions, and adapt its path toward a goal, which makes it useful for chargeback documentation, refund review, and routing decisions. The governance issue is that these systems create decisions at runtime, so their outputs must be traceable, bounded, and reviewable. If the system cannot explain why it acted or how it used signals, teams cannot safely rely on it for customer-impacting workflows.
Practical implication: Require audit trails, human oversight, and KPI-linked validation before agentic workflows touch money movement.
Why identity resolution matters more than single-point authentication
Identity resolution connects the same customer across login, payment, device, and loyalty activity. That matters because account takeover is rarely visible in a single event. Attackers often look legitimate at one step and fraudulent at another. When fraud teams only score a login, they miss the broader pattern. When they only review a refund, they miss the earlier compromise signal. The technical challenge is correlating identity, behaviour, and device context without adding so much friction that legitimate customers abandon the journey.
Practical implication: Build fraud decisions around joined-up identity signals, not isolated authentication events.
Threat narrative
Attacker objective: The attacker wants to monetise compromised customer accounts while avoiding detection long enough to extract value through purchases, refunds, or loyalty abuse.
- Entry begins with credential stuffing or phishing against customer accounts, giving attackers access to accounts that still rely on passwords or weak recovery paths.
- Credential access is sustained when the attacker reuses the authenticated session or enrolls a trusted device, bypassing basic login friction.
- Impact follows when the account is used for fraud, false refunds, or loyalty abuse, which damages both direct revenue and trust in downstream risk models.
Breaches seen in the wild
- Twilio 0ktapus breach 2022: SMS phishing of employees exposed 209 Twilio customers and 1,900 Signal users, part of the 0ktapus campaign against 130+ firms.
Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.
NHI Mgmt Group analysis
Passkeys are reducing password risk, but they are not eliminating identity trust gaps. The article shows that phishing-resistant authentication is becoming practical, yet merchants still need device trust, recovery governance, and journey-wide correlation. That is the important distinction: a stronger login does not automatically create a trustworthy customer identity. Practitioners should treat passkeys as one control in a larger identity trust model, not as the model itself.
Agentic AI introduces a governance problem because runtime action now sits inside customer operations. When systems can draft chargeback evidence or route reviews, the control question shifts from accuracy alone to bounded authority, traceability, and reviewability. That makes AI governance and fraud governance converge in the same workflow. The practitioner takeaway is that every customer-impacting agent needs explicit limits, logs, and oversight tied to business metrics.
Identity resolution is becoming the core control concept in digital commerce. Merchants cannot manage account takeover, refund abuse, and loyalty fraud as separate problems if the same customer journey spans them all. This is a verification trust gap, not just an authentication gap. Teams should design controls that connect login, device, and payment behaviour so fraud decisions reflect the whole identity story.
Digital commerce security is now a revenue architecture question, not just a fraud question. The article’s core message is that trust has to be measured in operational terms, including approval rates, false declines, and customer retention. That aligns with the broader move toward identity governance as a business control plane. Practitioners should expect fraud, IAM, and payments teams to share accountability for the same trust outcomes.
From our research library:
- eBay's passkey data shows 55-60% of passkey adoption happens on mobile, against around 20% on desktop.
What this signals
Identity trust will become a control layer in its own right. As passkeys spread and agentic AI starts handling more operational work, the programme risk shifts from isolated login hardening to end-to-end trust governance. Practitioners should expect more pressure to show how authentication, device trust, and customer identity resolution work together across the journey.
Merchant fraud teams will increasingly need identity architecture thinking. The old separation between authentication, fraud scoring, and payments operations is breaking down. Teams that cannot connect these disciplines will struggle to explain both revenue leakage and false-positive friction, especially when AI starts influencing customer decisions.
Account takeover is now a measurement problem as much as a prevention problem. Security teams need to know which controls reduce compromise, which ones protect conversion, and which ones create blind spots in recovery or refund flows. The winners will be the organisations that treat trust as a measurable business control, not a marketing promise.
For practitioners
- Expand phishing-resistant authentication coverage Prioritise returning-customer logins, high-value transactions, and account recovery flows for passkey rollout, with fallback paths tightly controlled.
- Bind authentication to device trust signals Combine passkeys with device fingerprinting, trusted-device policies, and step-up checks when behaviour diverges from the normal customer profile.
- Constrain agentic AI to bounded fraud workflows Use human review, auditable decision trails, and clear approval thresholds for chargeback and refund automation so agent actions stay explainable.
- Unify identity signals across the customer journey Correlate login, transaction, refund, and loyalty activity so account takeover detection does not depend on one isolated control point.
Key takeaways
- Passkeys strengthen authentication, but merchants still need device trust, recovery controls, and fraud correlation to make customer identity reliable.
- Agentic AI can support chargeback and refund workflows, but only if its authority is bounded and every action is auditable.
- Account takeover is now a revenue issue as well as a security issue, which makes shared identity metrics essential.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Agentic AI Top 10 addresses the attack and risk surface, while NIST SP 800-63, NIST AI RMF and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST SP 800-63 | SP 800-63B — Authentication | The article cites passkeys and phishing-resistant authentication as the main control change. |
| SP 800-63C — Federation | Identity trust across the customer journey depends on how authentication is bound and asserted. | |
| Recommendation — Adopt phishing-resistant authenticators and govern fallback authentication paths under SP 800-63B. Align federation and trust assertions so downstream systems can rely on authenticated identity signals. | ||
| NIST AI RMF | GOVERN — AI Governance and Accountability | Agentic AI used in fraud workflows needs governance, accountability, and oversight. |
| Recommendation — Set governance ownership and oversight for agentic AI before it is allowed to automate customer-impacting decisions. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agentic systems acting in commerce workflows can overstep intended authority if not bounded. |
| Recommendation — Constrain agent privileges and review runtime decisions to prevent identity and privilege abuse. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Customer authentication and step-up controls map to access permission governance. |
| Recommendation — Review entitlement paths and step-up controls so only trusted sessions reach high-risk actions. | ||
Key terms
- Passkey: A passkey is a passwordless credential based on public key cryptography. A private key stays on the user’s device, while a public key is stored by the service. During login, the device signs a challenge after local unlock, which reduces phishing and eliminates shared secret reuse.
- Account Takeover: Account takeover is unauthorized use of a legitimate account after an attacker obtains valid access through stolen credentials, tokens, or trusted integrations. The key security problem is that the resulting activity often looks normal to logs and controls, which makes containment and attribution harder than in a forced-entry breach.
- Agentic AI: Autonomous AI systems capable of planning, deciding, and taking actions, including calling APIs, writing code, and orchestrating other agents, with minimal human oversight. Agentic AI introduces new NHI risks as agents must authenticate to external services.
- Identity Resolution: Identity resolution is the correlation step that determines whether multiple accounts belong to the same person or accountable role. It combines identifiers, context, and system-specific attributes to reduce false splits and missed matches, which is what makes governance outputs dependable rather than approximate.
Deepen your knowledge
NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course. Explore nhimg.org for resources that connect identity governance to the broader security disciplines your programme depends on.
Published by the NHIMG editorial team on July 11, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org