At NYPAY’s Feb. 17 virtual event, Lee Kyriacou kept us out of the usual panel rut by pushing a simple idea: stop talking about scams in the abstract and walk through what happens in real life.
So that’s what we did.
We used real-world scenarios, asked the audience to vote “allow” or “deny,” then talked through what information would have made the decision easier, or at least less wrong. The format turned out to be the point: in scam prevention, the hard part isn’t knowing scams exist. The hard part is deciding what to do in the moment, with incomplete information, when the customer is involved and the money can move fast.
The panel brought four different lenses:
- Lee Kyriacou (moderator) kept the conversation grounded in practical decision-making
- Amanda Balmer shared several scenarios drawn from real situations institutions deal with every day
- Uri Rivner focused on behavioral signals and how coercion can show up in ways traditional fraud tools miss
- Cihat Fitzgerald brought an “upstream” view, stopping scam activity earlier in the chain, including at the merchant layer
- Peter Tapling connected the dots on incentives, tradeoffs, and why this problem is so hard to “policy” our way out of
Authentication isn’t the finish line
One theme that kept coming up: a customer can be fully authenticated and still be in trouble. People being coached or pressured can confirm details, pass step-up checks, and insist everything is fine; because they believe it, or because they’re afraid.
That’s why so many scam cases don’t show up as classic fraud. The “bad guy” isn’t stealing credentials and sneaking in. The bad guy is steering the customer.
The best control is often… a good question
Across the scenarios, the turning point was often not a fancy model. It was a well-timed question that forced a pause:
- What’s this payment for?
- Who is this person to you?
- Why does this need to happen today?
- What happens if we wait 30 minutes?
Scams thrive on urgency. Even small delays and a few calm questions can change the customer’s mindset from “act now” to “explain why.”
Receiver signals and context matter more than we admit
We also talked about how limited banks often are when they only see their own side of the transaction. If you don’t have good visibility into the receiving account or the receiving party’s history, you’re guessing, especially in faster payments, where guessing wrong can be permanent.
This is one reason receiver-side signals (where available) can be so powerful. It’s also why collecting and using customer-provided context, in a structured way, can help separate “looks weird but fine” from “looks fine but dangerous.”
Behavioral signals can help; but they’re not magic
Uri made the case that behavioral signals (things like hesitation, distraction, and patterns in how someone navigates a payment flow) can sometimes reveal coaching or stress. That’s especially relevant in scams, where the customer is present and “authorized” activity can still be harmful.
At the same time, everyone agreed on the obvious-but-important point: behavioral signals need context. People can look “weird” online for totally legitimate reasons (helping a parent, using accessibility tools, multitasking). The goal is to treat this behavior as a reason to ask better questions.
There’s no clean answer. and that’s the honest part
The format worked because it exposed reality: there are situations where you can’t be 100% sure.
You’re balancing customer protection, speed, reversibility, and trust. And when institutions get it wrong,
the damage can happen on both sides:
- scams that go through because everything looked “authorized”
- legitimate payments that get blocked with no explanation and no human recourse
In other words, the failure isn’t always detection. Often it’s the handoff. from signal to action, from system to human, from policy to what customer support can actually do.
Bottom line
This session left us with a pretty clear takeaway: scam prevention isn’t one thing. It’s a chain of decisions. Better outcomes come from combining:
- smarter signals (including behavioral and receiver signals where possible)
- better context collection (“why are you paying?”)
- an operating model that can actually act on what the system flags
And if we’re honest, we need more conversations like this: the kind where people are willing to sit in the gray area and talk about what really happens.