MRC Vegas Was Buzzing About Agentic Commerce…But Here’s A Reality Check.

What the Data Actually Shows About AI-Driven Transactions in 2026

Every year, MRC Vegas delivers no shortage of energy: packed sessions, hallway debates, and a long list of “next big things” the industry is on the verge of adopting. This year was no different. If anything, one topic dominated more conversations than most: agentic commerce.

By day three, though, something interesting happened. People weren’t energized by the topic anymore. They were tired of it.

That’s usually a signal worth paying attention to.

Why have FOMO about something nobody’s doing?

During one of my sessions at MRC, I opened with a simple question: How many of you are tired of hearing about agentic commerce? Most of the room raised their hands. Then I followed with the line that seemed to resonate the most:

“If you bet red on the roulette table 19 times in a row and win, that’s roughly the probability of a single fully agentic transaction today – about one in 500,000.” 

We were in Vegas, so the metaphor landed. But my point was serious: that estimate isn’t theoretical. It’s based on what we see across Accertify’s client consortium data (10 billion transactions in 2025 alone). With a dedicated engineering team focused on device and automation research, we can identify AI‑assisted activity – alongside other good and bad bots – and measure how often (or rarely, in this case) those agents actually make it all the way to checkout. 

We’re treating an extremely rare event like it’s already the norm. That mismatch is what creates unnecessary fear, rushed roadmaps, and misplaced priorities. So my take coming out of MRC is simple: why have FOMO about a party no one is actually attending?

What’s real today, and what isn’t

To be clear, this isn’t a dismissal of agentic commerce as a long-term concept. It’s a call for perspective.

What is happening today is that shoppers are increasingly using AI-driven- tools for search and comparison. Roughly 50% of consumers already rely on some form of agent assistance to evaluate options, compare prices, or narrow choices.

What isn’t happening at any meaningful scale is fully autonomous checkout. When it comes time to complete the transaction, people still want to be in control. That gap matters – especially for fraud and identity teams.

We aren’t the only ones seeing this gap. Walmart recently shut down its agentic commerce implementation with OpenAI, and OpenAI itself has scaled back native instant checkout inside ChatGPT – shifting purchases back to merchant‑controlled experiences.

My point here isn’t that agentic commerce won’t matter – it’s that its impact is arriving far more unevenly than the hype suggests.

From a fraud perspective, the fundamentals haven’t changed. The behaviors, incentives, and attack patterns are familiar. What’s changing is the medium, not the underlying mechanics.

The bigger risk isn’t agentic commerce. It’s AI-accelerated- fraud.

Here’s the part I think deserves more attention.

While the industry debates a future state that’s statistically rare, fraudsters are already using AI tools to scale what works today. Faster scripting. More convincing phishing. Synthetic identity creation at unprecedented speed. Automation that reduces effort and increases volume.

That’s the real acceleration curve we should be planning for.

“According to Accertify’s analysis of 3.878 billion transactions from December 1, 2025 to March 15, 2026, only 7,741 were agentic.”
(Source: Accertify client data, 12/1/25 to 3/15/26)

If you’re responsible for fraud prevention, identity confidence, or trust, you don’t need to overhaul everything for a checkout model that barely exists yet. You need to strengthen your ability to detect anomalies, intent, and behavioral patterns across channels, regardless of whether the actor is human or assisted by an agent.

Takeaways from MRC

After speaking with customers, prospects, and peers at MRC, a few themes stood out clearly:

• Continuous decisioning across the customer journey is paramount. The ability to confidently approve good transactions depends on evaluating identity, intent, and behavior over time – not reacting to single moments in isolation.
• Layered decisioning still wins. No single model explains fraud in isolation. Overlap is where accuracy improves and friction drops.
• Identity confidence matters more than interfaces. A slick front end means nothing if the back-end intelligence can’t keep up with evolving behavior.

The irony is that while everyone is worried about agentic commerce, the tools required to manage it look a lot like the tools many merchants are already investing in today.

The bottom line

Agentic commerce will evolve. One day, it may even become common. But right now, it’s over-represented in the conversation relative to its actual impact.

Coming out of MRC Vegas, my advice is this: Stay curious, but don’t let hype distract you. Stay grounded. There are enough real problems to solve NOW without borrowing anxiety from a future that hasn’t quite arrived yet. 

Keep fraud strategy grounded as AI accelerates

If you want to go deeper on what AI changes in fraud and commerce, and what it does not, join us May 11–14, 2026 in St. Petersburg, FL for the Accertify Global Summit. The program is designed for practical takeaways and peer discussion across fraud, identity, and digital commerce teams. Attendance is open to Accertify clients and select others. My session, “Agentic AI, Fraud, and Commerce: Don’t Panic!”, will focus on how to stay grounded, prioritize the right risks, and build roadmaps that reflect what is happening now.

Explore the agenda and request to attend the Accertify Global Summit here.