Riskified webinar key takeaways – Is AI-driven shopping putting retailers at risk?

Diana Vorniceanu, Senior Editor at The Paypers, discusses the key takeaways of Riskified’s ‘Is AI-driven shopping putting retailers at risk?’ webinar.

In 2025, agentic commerce rewrote the rules of online shopping and, in the process, opened up new discussions regarding risk and liability. Data from Adobe shows that, for the 2025 holiday season (1 November to 31 December), online consumer spending is anticipated to reach USD 253.4 billion, up 5.3% YoY, with AI traffic to retail sites rising to 760%. The numbers tell a compelling story: AI agents aren’t coming, they are already here. 

In a webinar hosted by Mélisande Mual, Publisher and Managing Director at The Paypers, Jeff Otto, CMO of Riskified and Kevin Luh, Partner at Deloitte, sat down to answer the question: is AI-driven shopping putting retailers at risk? The discussion highlighted how agentic commerce is transforming online payments and changing the risk, liability, and conversion playbooks in the process.  

Here are the key takeaways of the webinar.

Fraudsters are the fastest adopters of new technology

Jeff Otto started the discussion with a stark reality check: a screenshot from a dark web blog where fraudsters share step-by-step instructions on how to exploit a newly launched AI shopping platform. The implications are clear: while merchants are still debating whether to integrate agentic commerce, fraudsters are already reverse-engineering it.

He also highlighted that fraud service offerings are becoming more diversified. Returns-as-a-service, deepfake-as-a-service and other service kits of this type can be easily bought online, and they allow fraud rings to specialise and scale at a fast pace. 

Perhaps one of the most concerning fraud-related trends presented is the fact that fraudsters are turning to targeting entire industries rather than solely individual businesses. Once bad actors find a vulnerability, they systematically replicate attacks across merchants from the same vertical. 

For the payments industry in general, and merchants in particular, the important question that emerges is: how do you differentiate the malicious AI agents from good customers using AI shopping agents? The question grows more pressing as AI-enabled shopping matures from consumers getting curated shopping advice from AI agents to fully delegated AI shopping buyer agents, which increases the risk of fraud. 

Consumer adoption of agentic shopping is outpacing merchant readiness

In October 2025, Riskified conducted a global survey of 5,400 consumers, aged 18 to 64, that showcases consumer sentiment on using AI for shopping. The results are conclusive: people are growing more comfortable with the idea of using AI agents and letting them transact on their behalf. According to the data: 

• 73% of respondents use AI assistants like ChatGPT to shop
• 58% plan to use AI for holiday gift shopping
• 70% are comfortable letting AI transact on their behalf
• Payment security is the number one concern about agentic commerce 

The merchant side tells a different story. According to a Deloitte survey of 39 North American enterprise merchants – spanning C-suite, VP, and director roles across ecommerce, retail, and hospitality – 69% of respondents have already experienced AI-enabled fraud, yet only 3% feel well prepared to address it. 

Kevin Luh, Partner at Deloitte, framed the stakes by referring to the bigger picture: agentic AI has lowered the barrier to sophisticated attacks, but the damage extends well beyond financial losses. When left unaddressed, AI-enabled fraud undermines a merchant’s growth opportunity and impacts the customer experience and loyalty.  

However, the same Deloitte survey shows that merchants are not standing still, with 95% of the respondents saying that they plan to adapt their strategy for agentic AI. Turning that intent into execution could, however, be challenging, as 52% of merchants say budgets, skill sets, and technology are current pain points they are facing. 

Tokenized transactions raise concerns for issuers

Agent-initiated payments also face challenges on the issuer side, as issuers are hesitant to approve tokenized transactions, like those from Apple Pay. In these early days for agentic commerce, Kevin Luh explained that issuers face three major concerns:

1. The visibility problem: issuers traditionally rely on digital telemetry data via 3DS and other channels to understand customer behaviours. Agent-initiated payments introduce new data limitations.
2. The lack of a unified industry standard: frameworks for agent-initiated payments are still evolving, with no unified industry standard or best practice currently in place.
3. The liability question: liability is not a problem that can be managed by just investing in technology. A systematic approach is needed to address liability across current and future payment ecosystems. 

The competing protocols and their distinct visions for agentic payments

The webinar highlighted the three distinct protocols that have emerged to enable agentic commerce: OpenAI and Stripe’s Agentic Commerce Protocol, the Visa Intelligent Commerce, and Google’s Agent Payments Protocol, each with its own priorities and approaches. Kevin Luh offered a step-by-step analysis of the three protocols and explained how each of them tackles the task of demonstrating user authorisation and their vulnerabilities in the face of bad actors. 

As the industry is still racing to define protocols for agentic commerce, for merchants, this creates difficulties in deciding which of the frameworks to support, especially as the gaps in the systems are already visible.

Jeff Otto, CMO of Riskified, proved this with a test his team performed to assess the current vulnerabilities in Stripe’s ACP. They submitted a deliberately suspicious order using a new fake email, a debit card with an incorrect billing address, a VPN to shift their IP to another location, and a shipping address that didn’t match the billing address. The transaction was approved without a step-up challenge. The Riskified team then ran a second order through that same merchant’s ecommerce storefront. This time, the order was declined, and the account was blocked. 

What’s next for agentic shopping in 2026

Looking ahead to 2026, agentic checkout volumes are expected to rise. As evidence for this, Jeff Otto cited evidence from Amazon, which showed that the number of customers who used Rufus and bought something was 100% higher compared with baseline traffic during Black Friday, while shopping sessions where people did not use the AI agent and made a purchase only rose 20%. 

As agentic commerce protocols are still being perfected, for the payments industry, success in this new era of AI-enabled ecommerce will require collaboration and the rethinking of previously held assumptions regarding concepts like identity, authorisation, and trust. In light of the significant increase in AI-driven traffic during 2025’s Cyber Week, merchants need to start adapting today and invest in capabilities that can help them distinguish between good and bad bots, all while offering consumers the frictionless experience that they have grown accustomed to. 

This article only covers the main talking points, but there is more to learn about the topic by watching the full recording, which includes live audience polls on how many merchants are already testing or piloting agentic protocols and whether they have started piloting or already implemented their own custom shopping agents. To get the full scoop and see how other merchants are approaching agentic commerce, watch the full webinar recording here. 

About the author

 

Diana Vorniceanu is a Senior Editor at The Paypers. She has an extensive background in content creation and is a graduate of Foreign Languages and Literature studies, currently specialising in payments and ecommerce. She strives to bring forward the latest trends for our readers, while investigating the ever-evolving landscape of cross-border payments, global ecommerce, payments for marketplaces and online platforms, and emerging technologies across the globe. You can reach Diana via email at diana@thepaypers.com or on LinkedIn.

 

About Riskified

Riskified (NYSE:RSKD) empowers businesses to unleash ecommerce growth by outsmarting risk. Many of the world’s biggest brands and publicly traded companies selling online rely on Riskified for guaranteed protection against chargebacks, to fight fraud and policy abuse at scale, and to improve customer retention. Developed and managed by the largest team of ecommerce risk analysts, data scientists and researchers, Riskified’s AI-powered fraud and risk intelligence platform analyses the individual behind each interaction to provide real-time decisions and robust identity-based insights. Learn more at riskified.com.