Skip to content

Battling bots: How AI makes ecommerce more secure and profitable

Read “The State of Commerce 2023” report

“The State of Commerce 2023” report

Cover of Signifyd's State of Commerce 2023 report

It’s a wonder we’re not all running from bots in our nightmares. As the fascination with AI, generative in particular, has reached new heights advanced automation is often presented according to what it can do to us — steal our identities, steal our money, steal our intellectual rigor. 

It was a point Axios cybersecurity reporter Sam Sabin made at Collision 2023 when she sat down at the Toronto techfest with Signifyd CEO Raj Ramanand and Tailscale CEO Avery Pennarun to talk about bots in the real world.

It’s AI vs. AI in protecting the enterprise

It turns out, of course, that bots can do a lot of good, particularly when it comes to protecting businesses from some of the very stuff of our dreams.

“That’s where AI is starting to get interesting,” Ramanand said from the Collision stage. “It’s not a human versus AI anymore. Companies like ours and others have built AI products for a while. It’s now AI versus AI.

Watch the entire “How to turn bots from security risk to security advantage” fireside chat from Collision 2023. The discussion includes Signifyd CEO and co-founder Raj Ramanand, Tailscale CEO Avery Pennarun and Axios Cybersecurity reporter Sam Sabin.


Collision is an annual summit on tech, innovation and the future that organizers said drew 40,000 attendees this year along with more than 1,500 companies and hundreds of investors. Ramanand was appearing for the second time at the conference, which not surprisingly was dominated by discussions around AI in its many forms. 

AI bots have been fighting crime for years

Bots have been employed in a life of crime for some years now — for instance making rapid card-testing attacks and product scalping possible in the ecommerce world. But with generative AI on the rise, so too is the potential for automated wrongdoing. 

Ramanand, for instance, talked about the potential for criminal rings to bring phishing and other social engineering to a whole new level. It’s no longer about the grammatically challenged note from a supposed Nigerian prince asking for money, Ramanand said. Almost no one would fall for that sort of email today.

ChatGPT ups the game when it comes to phishing

“You can go into chat GPT and ask it to write that same email sitting in a cafe in Nigeria —  even if your English is not going to be the best — and you can pull off the most beautiful looking email that can be sent to my mom and she can look at it and say, ‘Well, this looks real. This might be real.’ And she may click on it.”

All of which is interesting to ecommerce businesses that must constantly and quickly determine which orders coming their way are legitimate and which are fraudulent. Phishing emails, you see, provides the raw material — personally identifiable information — for committing online fraud.

AI for good is gaining sophistication

“The end to this game is not so much to get them to click on it as much as it is to get their financial credentials,” Ramanand said. “Everybody is moving to stored value credits. They want you to not do guest checkouts. They want you to be creating accounts. If I ask you to log in and give me that information, now I have your login and password, or I have your credit card number and I can, at scale, as Avery was talking about, commit fraud.”  

But here’s the thing — and it’s something that Ramanand and Pennarun talked about before the crowd in Toronto — companies like Signifyd and Tailscale have automation on their side as well. And it’s also getting more sophisticated by the day. 

Ramanand walked through the challenge online retailers have protecting themselves from fraud and how the combination of artificial intelligence and data can solve it. He said that historically merchants have added friction to the buying process as a means to avoid filling fraudulent orders. 

How do smart machines make for a better customer experience?

But it turns out many typically overdo it — turning away 10% or more of orders when the percentage of fraudulent orders placed is closer to 1%. 

“If you translate that to a physical store experience,” Ramanand said, “that’s like turning away, for every 10 new users that come into your store, you turn away one good user. And that’s a pretty big problem if you’re doing that online because at scale that’s a lot, a lot, of lost revenue. And so the most valuable application from an AI perspective is to help retailers leverage the power of data to identify what is good and what is bad.” 

In fact, Ramanand said, online retailers turn away about $400 billion in good orders in an overzealous effort to avoid fraudulent orders. 

How does automated fraud detection add billions to the economy?

“From an economy standpoint, that’s a major impact,” he said. “Part of where AI can be massive and continues to be in companies like Signifyd is to be able to go out and say, ‘We can bring that back into the economy” and by saying we can reduce some of the losses, take away the liability and convert 400 extra billion, all with the power of AI. That’s a massively valuable resource for us to be able to pursue.” 

All of which puts in perspective how good bots can keep our nightmares at bay.

Video courtesy of Collision. Photo courtesy of Tailscale.

Looking to unleash the power of AI to reduce your fraud protection costs? We can help.

Mike Cassidy

Mike Cassidy

Mike is the head of storytelling at Signifyd. A former journalist and a retail geek, he covers ecommerce and the way technology is transforming digital commerce. Contact him at [email protected].