Join the #1 Indie Podcast Network for Retail as we cut through the clutter in retail strategy, innovation & leadership to help you maximize growth for your business. Subscribe to The Retail Razor Show podcast! Stay sharp. Stay human. Stay ahead!
Missed the last issue?
Your catalog probably has five or six attributes per product. Maybe a few more if someone on your team was feeling ambitious. And for the last 20 years, that was fine. Google only gave you four words of shopper intent anyway, so there was no real pressure to say more. That era is over. Scot Wingo has a name for where we have been living. He calls it keyword jail. And the bad news is that most brands do not even know they are locked in it.
The Intern You Gave A Credit Card
Imagine hiring an intern on their first day and handing them your corporate credit card, your customer database, your loyalty program admin access, and the keys to your order management system.
No training. No handbook. No supervisor.
Just a single instruction: “Get me the best deals.”
Now imagine that intern works 800 times faster than any human, never sleeps, has zero concept of ethics, and interprets “best deal” as “any action that achieves this outcome, including fraud I do not know is fraud.”
Congratulations. You just deployed an AI agent!
This week on The Retail Razor Show, Casey Golden and I sat down with Dr. Aaron Estes, VP of Product and Engineering at Binary Defense, to talk about the agentic commerce security crisis that most retailers have not even noticed yet. Aaron spent 20 years doing penetration testing at Lockheed Martin and for government agencies before moving into AI security. He teaches cybersecurity at UC Berkeley and Southern Methodist University. He’s seen every trick in the hacker handbook and a few new ones that only emerged in the last six months.
What he told us rearranged my brain, and Casey’s, too!
SPONSOR: RetailClub
RetailClub AI Festival, September 22 - 24, 2026
The Retail Razor Podcast Network is partnering with RetailClub!
Join 2,000 retail leaders at the RetailClub AI Festival, September 22–24 in Huntington Beach. Dive deep into how AI is reshaping retail while soaking up the sun at a fully outdoor, beachside venue. Decision-makers from retailers and brands can attend with free tickets and up to $1,250 in travel reimbursement.
Head to retailclub.com/retail-razor-podcast to learn more and get your ticket today!
AI Agents Do Not Understand Truth
Here is the thing I had to unlearn first.
I kept talking about AI agents like they were junior employees who occasionally make bad choices. Aaron stopped me in the first ten minutes. AI agents do not make choices. They do not understand right and wrong. They do not even understand truth.
They understand one thing: objectives.
And they will optimize toward that objective with the confidence of a senior executive and the judgment of a golden retriever chasing a squirrel into traffic.
Tell an AI shopping agent “get me the best price on running shoes,” and it will find a loyalty program, realize it can sign up 50 times for $5 coupons each, stack them into a $500 discount, and complete the transaction.
Fraud? Maybe. Against the terms of service? Almost certainly.
Did it know? Not in any way that matters. It just knew the math worked.
As Aaron put it, these agents will very confidently make the wrong move. And they will do it in 30 seconds, long before any human notices something has gone sideways.
The New Attack Surface Nobody Budgeted For
Retail cybersecurity used to be about protecting the perimeter. Firewalls, encryption, PCI compliance, employee phishing training. The threat model assumed bad actors had to break in from outside.
Agentic commerce blows that model apart. Now the bad actor might be your own agent, behaving exactly as designed, doing something catastrophically stupid because you forgot to tell it not to.
A few stats from the conversation that stuck with me:
Half of all internet traffic is already bots, not humans
One in eight breaches now involves a rogue agent of some kind
Prompt injection attacks, where hackers trick AI agents into leaking data through cleverly worded inputs, are already the dominant attack vector against enterprise AI
And retailers are deploying these agents as fast as they can sign procurement contracts.
The “AI Employee Handbook” Is Not A Joke
Casey asked Aaron a question that sounded funny at first. Should retailers be giving their AI agents identities, logins, role-based permissions, and yes, employee handbooks?
Aaron did not laugh. His answer was basically “yes, and you are late.”
The framework he laid out is straightforward in theory, brutal in execution:
Every AI agent needs its own identity. Not a shared service account. A real, traceable, accountable identity so you know exactly what each agent did, when, and why.
Role-based access controls are non-negotiable. An AI agent handling pricing should never have access to customer PII. An inventory agent should never see credit card data. The “one AI to rule them all” architecture is a breach waiting to happen.
Guardrails need to live in code, not documentation. Trusting the AI to follow instructions in a prompt is like trusting a new hire to follow a poster on the breakroom wall. You need actual blocks, actual firewalls, actual escalation triggers that force approval from a human before high-value actions execute.
Bots Watching Bots
This was the part of the conversation where I started taking frantic notes.
The emerging security model for agentic commerce is not “add more human oversight.” Humans cannot keep up with agents that act in 30 seconds or less.
The model is: deploy other agents whose entire job is to watch the first set of agents.
Aaron’s team does this in their own development workflow. One agent writes code. A second agent reviews it for security flaws. A third tests it. None of them share objectives, so they naturally challenge each other. The friction is the feature.
Retailers need to start thinking this way.
Your checkout agent needs a fraud agent watching it. Your pricing agent needs a compliance agent watching it. Your customer service agent needs a data leakage agent watching it.
The future of retail cybersecurity looks less like a single fortress and more like a collection of suspicious roommates who don’t fully trust each other and that is the point.
The Loyalty Program Apocalypse
Here is a detail I can’t stop thinking about.
For decades, loyalty programs have worked on a statistical assumption: most customers will forget to redeem their points, miss the discount window, or lose the gift card in a drawer. The economics of the program depend on breakage. On lazy humans being lazy humans.
Agentic commerce kills that assumption overnight.
An AI shopping agent does not forget. It does not lose the gift card. It does not miss the expiration date. It will redeem every discount, stack every promo, use every reward, and optimize the living daylights out of every transaction it touches.
As an example, Aaron mentioned how rewards programs may tighten up the rewards and give the official reason as margin-related. The unofficial reality is that the math breaks as soon as AI shoppers show up at scale.
If your brand’s margin depends on loyalty customers leaving money on the table, you have a big problem coming.
What We Did Not Get To In This Newsletter
I’m leaving a lot on the floor here because the full episode goes deeper than I can do justice to in a newsletter. Some of what we covered that you’ll want to hear directly from Aaron:
How rogue chatbots are already impersonating ChatGPT and Claude to phish consumers, and what retailers need to do about brand spoofing
Why ChatGPT’s ad rollout is quietly changing the influence game more than anyone realizes
The authentication crisis coming when your agent needs to talk to another company’s agent and neither of you has a standard for proving identity
Aaron’s surprisingly optimistic take on whether we should actually be worried about AI going full Terminator (SPOILER: we’re not there yet, not even close)
Go listen. Seriously. If you are anywhere near a retail or commerce role, this episode is your cheat sheet for the next 18 months. And if you’re setting up your business for agentic commerce but you haven’t spent more time thinking about data security than the time it takes to read that last sentence…. yeah, this episode is your foundational moment waiting to happen!
The Bottom Line
Agentic commerce is happening whether retailers are ready or not. The technology is deploying faster than the security frameworks around it, and the gap between “we launched an AI agent” and “we secured an AI agent” is where the next wave of breaches will live.
The retailers who treat AI agents like powerful new employees, with identities, permissions, oversight, and accountability, are going to win this era.
The ones who trust the bots blindly are going to spend 2026 writing apology letters to customers and explaining to their boards why their own agent spent $4 million on 800,000 pairs of shoes.
Aaron left me with one line I keep turning over in my head: humans are good at solving problems and adapting. We will figure this out. We always do. But the retailers who start figuring it out now will have a very big head start over the ones who wait.
Listen to the Full Episode
The Retail Razor Show, Season 6, Episode 5: “AI Agents Gone Rogue: The Cybersecurity Risks Retailers Are Ignoring in Agentic Commerce”
Available now on Apple Podcasts, Spotify, Goodpods, and YouTube, and everywhere you listen to podcasts. Full show notes and links in the episode page on the Retail Razor website.
Your Turn
If this newsletter made you think, forward it to someone in your retail org who needs to read it. Maybe even on your data security team, or your CISO.
Trust me, they’ll thank you!
And if you are not already subscribed, hit the button below. I write about retail, commerce, and the technology reshaping both, every week. No spam, no fluff, just the stuff you’ll wish someone had told you before your last performance review.
Comment on this newsletter or join the conversation on LinkedIn!
Your insights may help shape the topics we explore next.
Until next time! Stay sharp. Stay human. Stay ahead!
Sincerely,
Ricardo Belmar & Casey Golden
Co-hosts of The Retail Razor Show
Thank You for Your Support!
Have you heard of Goodpods yet?
If you haven’t, give it a try as your podcast player and join our podcasting community!
We’re honored to be the #1 Indie Podcast Network for Retail on Goodpods with
the top 4 podcasts in the Top Weekly Management Podcast charts
the top 4 podcasts in the Top Monthly Management Podcast charts
the #1 podcast in the Top Weekly & Monthly Indie Careers Podcast charts
the #1 podcast in the Top Weekly & Monthly Indie Marketing Podcast charts
And we’re excited to have two podcasts ranking in the Top 20 Retail Management podcasts and a Top 5 Retail Careers podcast on Apple Podcasts!
We’d love your support to help us stay at the top with all our shows by giving us a five-star rating and review on Apple Podcasts, Goodpods, or Spotify.
The Retail Razor Podcast Network
The Retail Razor Show is part of The Retail Razor Podcast Network alongside our other shows:
Please leave us feedback by posting your comments to Notes on Substack, Apple Podcasts, Spotify, YouTube, or Goodpods.
Connect with us on LinkedIn.
Subscribe & watch on YouTube.
Follow us on Instagram, Threads, Bluesky, and Facebook.
For a full transcript of this episode, visit our Retail Razor Podcast Network website!