Tackle Consumer Scams with Innovation and Interdiction

While we wait for critical 2025 scam numbers from a number of key countries, we are staring at the damaging impact of the evil twin of Gen AI with devastating attack vectors.

‍

It is true that consumer scams begin well before the bank customer/victim sends money to the scammer. It is also true that digital platforms and telcos need to do more to stop these scams from even beginning. But the reality is that the customer’s financial institution is the very last point to help stop the scam and it is essential that the financial institution (FI) puts up a good fight to keep the customer’s funds.

But how does the FI do this and what tools exist to help? The good news is banks, and telcos, in many countries have deployed meaningful controls to help stop scams. And we are starting to see innovative Gen AI solutions to help prevent scams. Here are just a few of the new Gen AI solutions:

1. Several companies are using an AI bot to talk with scammers to tie up their time. This is so good that the bot can even get money mule account information.   
2. Westpac Bank in Australia developed an AI agent in 2025 to help fraud analysts when talking with potential victims.  “The clever tech can detect indications the customer may be about to transfer money to a scam, based on their phone conversation with the bank. It can even detect whether a customer is being coached, or told what to say, by a scammer in the background.”
3. Several telcos have developed AI solutions to help consumers detect scam messages.

To me one of the most interesting changes in the fight against consumer scams, is the change from focusing heavily on the transaction alert alone to adding serious controls around the customer interdiction once the alert is generated (or the customer walks into a branch to withdraw large amounts of cash or request a wire).  

I have had several discussions with Michelle Hilscher, PhD, with Deloitte Canada’s Behavioral Economics Unit.  She focuses on training bank staff in the proper techniques of interdicting with a customer that might be a scam victim. It requires training the staff on the psychology of scams so the staff can understand the customer’s emotional state. This customer interdiction is so important, because just getting an alert, as good as it may be, does not keep the money in the bank. It is only by talking with the customer in a meaningful and effective way, that the FI has a chance to stop the scam. Remember, the customer might have been psychologically groomed for months, and what chance does bank staff have to ‘break the spell’ if they are not well trained. One point Michelle makes is that once you have your staff interdiction training program in place, your work is not done. Be prepared to continually update as your staff learns from these interdictions.

So, one key point so far is to allocate money for both scam detection/alerting and customer interdiction. Maybe the funding should be 70-30 or 80-20.

As I continue to research Gen AI capabilities, I ask myself if I can’t have a psychologist with me on every customer interaction to ‘help break the spell’, what is the next best thing? This is where we should be expanding on what Westpac started last year. In 2026, it is not unrealistic to use Gen AI to be part of the interdiction with the customer when discussing an alert. The ability to use psychological background (almost having the ‘psychologist in a box’) in framing questions based on the customer responses is clearly the way to go. 

If I was running a consumer scam prevention strategy, I would want effective staff training and the ‘psychologist in a box’.  After all, I need to do everything to keep the money with the customer. I want to prevent the need for recovery discussions. The well-being of a customer not being scammed is so important. There continues to be way too many stories of victims losing from $80,000 to $2 million or more in romance and investment scams.

I predict that many banks beyond Westpac will start to deploy sophisticated customer interdiction solutions. And the interdiction solution, along with strong detection/alerting, will make a real difference. I am especially watching Singapore, New Zealand, Australia, and UK banks to see how they use AI agents with fraud analysts for both alerting and customer interdiction.

One important caveat is that AI agents need appropriate guardrails, which some call ‘guardian’ agents. The concept of AI agents is new, and proper security is required to prevent the AI agent from being corrupted. In an article by Mill Pond Research, it was mentioned that “Today, AI agents are provisioning infrastructure, answering customer support tickets, triaging alerts, approving transactions, writing production code, and so much more.” It goes on to say: “that every autonomous agent must be governed, audited, and attested just like a human user or machine workload.”

Summary

We are in the best and worst of times of fraud and scam management. The worst of times because we see a growing number of scam losses. In 2024, the US Federal Trade Commission estimated US consumers lost over $195.9 billion in scam losses per year. This is similar in other countries on a per capita basis. The best of times in that there are some truly innovative solutions and ideas to help prevent scam losses.  

• I am seeing what Apple and Google are doing to improve the mobile phone operating system to prevent scam calls and text messages.  
• UK’s Virgin Media O2 uses AI to detect and flag 1 billion suspected scam and spam calls.
• I am blown away by the success that several companies have had in getting money mule accounts using AI agents that clearly sound real to the scammer.  

I am talking with entrepreneurs setting up their own companies to solve the scam problem.  I see their energy and know that we will see even more solutions in 2026.