Read “The Global State of Fraud and Returns” report
The fraud and payments world convened in Las Vegas for the Merchant Risk Council’s annual conference and the energy was clear: We’re no longer just fighting “fraudsters” and “bad actors”; we are preparing for a fundamental shift in how commerce is conducted.
The intersection of AI, consumer behavior and regulatory frameworks will shape the future of commerce and fraud prevention.
We spent the week hosting and attending sessions and discussing with merchants how they can turn these complex challenges into growth opportunities. Here are the four biggest themes and takeaways from MRC Vegas 2026 in case you couldn’t make it:
1. Agentic commerce: The reality of today and preparing for tomorrow
It comes as no surprise that agentic commerce was a hot topic at MRC Vegas. As consumers increasingly delegate shopping tasks to AI, we’re shifting from a solely destination-based web to one that includes agentic activity in which AI agents assist in discovery, cart building and final checkout. While this offers unprecedented choice, it’s also creating a “signal blackout” for traditional fraud defenses.
When agents run in data centers, our traditional “trust stack” (i.e. IP reputation, device fingerprinting, cookies, etc.) disappears, and if we treat this automated traffic as malicious, we block good orders.
Greg Smith, principal software development engineer at Amazon, and Varun Kumar, Signifyd’s chief technical officer, discussed the challenges that merchant fraud teams are facing and provided guidance on how to get started building the systems to support an agentic shopping experience.
Key insights:
- New standards like the Agentic Commerce Protocol (ACP) and Universal Commerce Protocol (UCP) are essential for managing agent-involved transactions, allowing agents to pass delegate tokens to merchants, enhancing security and configurability compared to traditional methods.
- There is a predicted rise in payment-initiation fraud and “friendly fraud” as agents take a larger role in shopping.
- Acceptance of these agents is not uniform with varying willingness among consumers to trust AI agents across different age groups (Gen Z: 63%, Millennials: 72%, over 55: <20%).
For more on the current state of agentic commerce and realistic guidance, check out our agentic white paper.
2. The humans in the loop: Where efficiency meets complex problem solving
A major theme of the conference was the integration of AI and ML in fraud prevention. The consensus from fraud experts at World Market, Belk and Canadian Tire is that while AI certainly promotes efficiency, human oversight remains essential especially for complex situations.
How pairing humans and machines turns fraud teams into revenue drivers
By finding the right mix of machine learning and human intelligence, ecommerce fraud and risk teams can change their missions from being defense focused to a focus on optimizing revenue, World Market’s Director, Internal Audit Jackie Mossberger explains.
Key insights:
- AI and ML are necessary to handle increased transaction volumes and provide quick decisions that humans cannot match.
- AI models often lack the emotional intelligence and contextual understanding required to address complex cases.
- The role of fraud analysts is shifting from manual review to “decline recovery,” focusing on recapturing revenue from transactions that AI may have misjudged.
3. Calculating the “total cost of fraud”
At many organizations, fraud is still viewed through a narrow lens — usually as a loss that needs to be minimized at all costs. However, Wayfair’s Head of Data Science & TMP Functions Matthew Lampert, along with Signifyd’s Chief Customer Officer J. Bennett, challenged this view by advocating for a more holistic, business-aligned metric: the total cost of fraud.
Why calculating the “total” cost is critical
Measuring fraud simply by the number of chargebacks you receive is like looking at a balance sheet with half the entries missing. Identifying the total cost is essential because:
- It allows the fraud team to speak the same language as finance and growth teams, transforming the department from a “roadblock” into a strategic partner that optimizes the bottom line.
- It helps leaders decide where to “strategically apply friction” to risky behaviors without ruining the experience for good customers.
- It forces a choice between short-term “growth at all costs” and building a long-term, profitable business.
The challenge: Capturing the “invisible” data
The hardest part of this calculus isn’t the money lost to bad actors; it’s the profit not captured (PNC). This represents the lifetime value (LTV) of “false positives” — good customers who were incorrectly declined and never returned. Most retailers find this number incredibly difficult to pin down because it requires knowing exact margin profiles and acquisition costs, and it often takes months of deep data mining to validate the correlation.
To reach a tangible number, Lampert shared a formula that combines direct and indirect costs:
- Direct costs: Sum up the fees paid to external data partners, the fines charged for chargebacks and the net chargebacks remaining after successful recoveries.
- Indirect costs: Layer in your total operational expenses (human review, etc.) and the estimate for profit not captured.
- Triangulating PNC: Lampert suggested four data points to estimate false positives:
- Control groups: Periodically run a “dark pool” where you approve a small sample of high-risk orders to see how many actually result in fraud.
- Market benchmarking: Compare your rates against industry averages for false positives
- Customer support feedback: Use a specialized service team to review appeals from impacted customers to assess if they were legitimate.
- Model stack ranking: Look at your “grey area” declines (excluding “obvious fraud”) and approximate the percentage that matches your other three data points.
By unifying these elements into a single KPI, Wayfair can optimize for the best bottom-line outcome, making it simple to decide when to approve or decline by asking: “Which decision minimizes our total cost of fraud?”
4. Collaboration against sophisticated threats
Sessions on ticketing, entertainment and hospitality underscored that fraudsters operate as organized networks, requiring merchants to respond in kind.
Key insights:
- Modern fraud is often structured like a business, with “admins” (technical experts), “workers” (phishers), and “callers” (fake customer service).
- Fraudsters target multiple platforms simultaneously; therefore, sharing insights and trends across internal teams and even industry peers is a powerful way to enhance fraud prevention strategies and catch new attempts early.
- Effective detection relies on clean, consistent data and the use of behavioral tools to identify anomalies in the transaction process.
The combination of agentic commerce, the growing sophistication of fraud and the increasing ability of meaningful data to provide direction are sure signs that the world of fraud and risk is rapidly evolving. Industry professionals doubled down on that reality at MRC Vegas 2026 last week and came away with a playbook for a new era.
We cover a lot of these themes and challenges in our 2026 State of Commerce report. Whether you were able to make it to MRC Vegas or not, it’s a worthwhile read for the latest on how the economy is influencing consumer behavior, a realistic state of agentic commerce, guidance around returns optimization and much more.
Photo by Signifyd
Want to learn more about the ongoing transformation of commerce? Let’s talk.
