The Side of Agentic Commerce We're Not Talking About

I just finished writing a whitepaper called “The Agentic Shift: A Strategic Analysis of LLM Impact on the 2025 Holiday Shopping Season.” Eighteen pages of data, forecasts, and strategic implications. Adobe Analytics. Salesforce projections. Google’s “Buy for Me” infrastructure. The whole thing.

It’s solid work. I stand behind it.

But something started bothering me as I was writing it. A thread I kept pulling at and then tucking back in because it didn’t fit the narrative. The whitepaper tells a story of efficiency gains, conversion optimization, and competitive advantage. The data is real. The trends are real.

And yet.

There’s a version of this future that looks less like “shopping gets easier” and more like “retail becomes a trading floor.” I couldn’t quite articulate it while I was in the weeds of the research. But now that I’ve stepped back, I can’t stop thinking about it.

Where This Started

In Section 2.1 of the whitepaper, I defined the “Three Tiers of Agentic Capability”:

Informational Agents: “Find me the best TV under $500.”
Monitoring Agents: “Track this TV and alert me if the price drops.”
Transactional Agents: “Buy this TV if it drops below $450 at a retailer with 2-day shipping.”

When I wrote that third tier, I framed it as a convenience feature. A time-saver. The logical next step after search.

But re-reading it now, I see something else.

That’s a limit order.

Not metaphorically. Literally. The user is setting a price target, a venue constraint, and authorizing automated execution. That’s the same thing a trader does when they tell their broker: “Buy 100 shares of AAPL if it hits $150.”

I kept writing. Didn’t stop to interrogate it.

Then I got to Section 5.2, the Black Friday forecast:

“Millions of ‘Buy for me’ triggers set by users in previous weeks will fire simultaneously as prices hit their floor.”

I was describing a liquidity event. Cascading order fills at a price support level. Flash-crash dynamics applied to consumer goods.

And I still didn’t stop. Because the whitepaper needed to be about holiday shopping, not market structure theory.

But now I’m stopping.

The Question I Should Have Asked

Here’s what I didn’t write in the whitepaper:

If we’re building infrastructure that allows millions of agents to execute price-triggered purchases simultaneously, are we building a retail market—or a financial market?

The more I sit with this, the more I think the answer is: both. And we’re only talking about the first one.

What Changes When Retail Becomes Trading

I’ve been trying to map financial market concepts onto what I described in the whitepaper. Not as metaphor—as direct analogs.

The Limit Order

In the whitepaper, I described consumers setting “bound parameters” for agents. Buy this espresso machine if it drops below $300. That’s not a shopping list. That’s an open order with a trigger price.

The whitepaper’s Section 1.2 makes this explicit:

“Consumers are using LLMs to ‘watch’ products starting in October, with instructions to execute purchases only when historical lows are breached during Black Friday week.”

That’s price-triggered execution. The same thing hedge funds do with algorithmic trading systems.

The Order Book

In the observability whitepaper I wrote earlier—“Measuring the Unowned Storefront”—I described an “Exposure API” where agents report their candidate_set[] with ranked products and reasons. Re-reading that now, I realize I was describing a bid-ask spread. Which products are being considered. At what prices. By which agents.

That’s an order book for physical goods.

The Flash Crash

The whitepaper mentions that Black Friday will stress-test agentic checkout infrastructure. I framed it as a server capacity issue.

But think about what actually happens:

Amazon drops a TV to $399—the trigger price for millions of waiting agents
All agents attempt simultaneous purchase
Inventory exhausts in milliseconds
Agents pivot to Walmart, overwhelming their systems
Walmart’s pricing algorithm raises price due to demand spike
Consumer agents see price rise, pause execution
Walmart inventory sits unsold while Amazon is sold out

That’s not a server problem. That’s a flash crash. Chaotic allocation driven by algorithmic feedback loops.

In 2010, the stock market dropped 9% in minutes because of similar dynamics. Finance responded with circuit breakers. Retail has none.

The Part I Didn’t Want to Write

Here’s where I start to feel friction.

The whitepaper is optimistic. It describes agentic commerce as a competitive advantage. Retailers who adopt it see 7x higher sales growth. Consumers save time. Everyone wins.

But there’s another reading.

Demand becomes synthetic.

If agents can be spun up programmatically, demand signals become manipulable. One bad actor with 10,000 agents can create fake purchase intent. Retailers see the signal, stock up on inventory that no one actually wants. Or worse—prices get driven up based on phantom demand.

In financial markets, this is called spoofing. It’s illegal. In retail, it’s just… Tuesday?

Prices lose stability.

The whitepaper mentions “algorithmic pricing wars between Amazon and Walmart.” I framed it as competition. But algorithmic pricing systems don’t always compete. Research suggests that given enough time, they learn that not competing is more profitable. They converge on higher prices without ever communicating.

That’s tacit collusion. Also illegal in finance. Unclear in retail.

Settlement doesn’t exist.

This is the one that really bothers me.

Financial markets have clearinghouses. The DTCC guarantees that when you buy a stock, the seller actually has it. Trades settle.

Retail has nothing like this.

The whitepaper’s Section 6.2 notes that “97% of retailers paid for clicks on unavailable products.” Ghost inventory. Phantom liquidity. When an agent tries to execute a “trade,” there’s no guarantee the product exists.

In finance, this would be like selling shares you don’t own without shorting. It would collapse the market. In retail, it’s the norm.

What I Used to Think vs. What I Think Now

When I started the whitepaper, I thought agentic commerce was a UX evolution. Better search. Faster checkout. The next iteration of the Amazon “1-Click” button.

Now I think it might be a market structure transformation. The same forces that turned stock trading from floor brokers to high-frequency algorithms—speed advantages, information asymmetry, arbitrage opportunities—are entering retail.

And we’re pretending it’s just “shopping, but smarter.”

The Infrastructure Gap

Here’s where the whitepaper’s optimism and my current skepticism collide.

The whitepaper describes the benefits of agentic infrastructure: observability, attribution, conversion. All real.

But it doesn’t describe the safeguards.

Financial markets have:

The SEC (market manipulation enforcement)
FINRA (broker-dealer rules)
Circuit breakers (trading halts during volatility)
Settlement guarantees (clearinghouses)
Position limits (caps on how much any entity can buy)

Agentic commerce has:

Nothing.

We’re building Ferrari engines on dirt roads. And celebrating the speed.

Who Builds the Guardrails?

This is the question I don’t have an answer to.

In finance, the crashes came first. The 1929 crash led to the SEC. The 2010 flash crash led to circuit breakers. Regulation followed failure.

Will retail follow the same path? Do we need a “Black Friday Flash Crash” before anyone takes market structure seriously?

Or is there a way to build the guardrails before the crash?

The whitepaper’s Section 8 has recommendations for retailers: real-time inventory accuracy, agent identity verification, observability infrastructure. Those are good starts. But they’re individual company decisions. There’s no industry standard. No neutral broker. No clearinghouse.

Maybe that’s the next whitepaper.

Or maybe someone else needs to write it. Someone with regulatory expertise, not just commerce architecture experience.

This Isn’t New. I’m Just Late.

The more I sat with this, the more I realized I was acting like retail invented something. We didn’t. We’re just the last industry to the party.

This “financialization” of buying decisions—algorithms replacing human negotiation, real-time bidding, automated execution—is already the standard operating model in advertising, logistics, and energy. Retail is importing mature architectures into the consumer world and pretending it’s innovation.

I should have seen the pattern earlier.

Advertising figured this out a decade ago.

When you load a webpage, an auction happens in milliseconds. Advertisers bid on your attention. No human negotiates that ad slot—it’s entirely bot-to-bot. They call it “Real-Time Bidding.”

Replace “ad slot” with “shampoo” and “advertiser” with “consumer agent,” and you have the exact model I described in the whitepaper.

The outcome? Efficiency and massive fraud. By 2024, over 20% of programmatic ad traffic was bots viewing ads bought by other bots. The “Sybil Attack” risk I worried about for retail? Advertising has been living with it for years.

Logistics got there too.

Uber Freight, Convoy (before it collapsed), and other “digital freight matching” platforms turned trucking into a spot market. An API posts a load. An algorithm accepts it instantly based on price, route, and timing. No human broker calling truckers to negotiate.

The result? Incredible volatility. Rates spike or crash in hours based on algorithmic demand signals. Traditional human brokers got margin-compressed out of existence.

Sound familiar?

B2B procurement has been doing this quietly.

Companies like Pactum AI provide autonomous negotiation bots. The bot chats with a vendor—human or machine—and closes deals without a human buyer ever reviewing them. Walmart used this for tail-end suppliers that human buyers didn’t have time to call.

So “agentic buying” already works at scale. For commodity items. In business contexts. The whitepaper is just describing the consumer-facing version.

Energy markets are the cautionary tale.

In deregulated markets like Texas, algorithms buy and sell electricity in 15-minute increments based on weather and usage data. The Griddy disaster in 2021 is what happens when you expose regular consumers to high-frequency pricing without circuit breakers.

When the grid froze and algorithms spiked prices to $9,000 per megawatt-hour, consumers on “real-time” plans were financially wiped out in days. They’d opted into algorithmic pricing without understanding the volatility exposure.

That’s the retail future I’m worried about. Not for sophisticated arbitrageurs—for regular people who don’t know they’re trading.

And then there’s Zillow.

Zillow tried to use algorithms to price and buy homes instantly. The algorithm failed to predict the velocity of the housing market downturn. It bought too many homes at too high a price because it was looking at trailing data.

They shut down the entire division and took a massive loss.

That’s the inventory risk. If a retailer’s algorithm—or a buying agent—misreads demand signals, it accumulates toxic assets. Unsellable inventory. Fast.

The Pattern I Should Have Seen

Industry	The “Asset”	The “Agent”	What Happened
Finance	Stocks	HFT Algorithms	Flash crashes, tight spreads
Advertising	Attention	Programmatic Bidders	Efficiency + massive fraud
Logistics	Truck Capacity	Digital Freight Matching	Volatility, margin compression
Energy	Kilowatts	Real-Time Pricing Bots	Griddy collapse, consumer harm
Real Estate	Homes	iBuying Algorithms	Zillow disaster
Retail	Consumer Goods	Shopping Agents	We’re about to find out

Retail is the final frontier because we deal with physical, heterogeneous goods—a shirt isn’t as standard as a kilowatt—and emotional humans. That’s why it’s taking longer.

But the playbook exists. We’re just running it on new inventory.

Where I’ve Landed—For Now

I’m not backing away from the whitepaper. The data is sound. The trends are real. Agentic commerce is happening, and retailers who ignore it will lose.

But I’m also not comfortable pretending this is just about efficiency.

We’re building a high-frequency trading floor inside a shopping mall. And right now, there are no referees.

The whitepaper describes the opportunity. This post is me wrestling with the risk.

I don’t know if the financial market parallels are perfect. Retail isn’t securities. Inventory isn’t shares. There are meaningful differences.

But the dynamics—speed advantages, information asymmetry, algorithmic feedback loops, settlement failures—those are converging.

And the more I look at what I wrote, the more I think we’re having the wrong conversation.

We’re talking about “AI shopping assistants” when we should be talking about market structure.

We’re talking about “conversion optimization” when we should be talking about settlement guarantees.

We’re talking about “agent adoption rates” when we should be talking about who regulates the agents.

The Uncomfortable Conclusion

The whitepaper ends with this line:

“The storefront was the interface. Now the agent is the interface. The storefront is just infrastructure.”

I meant it as a strategic insight. A call to action for retailers to adapt.

But there’s another reading.

If the storefront is just infrastructure, and agents are the interface, then who controls the agents controls commerce.

That’s not a shopping question.

That’s a market power question.

And I’m not sure we’re ready for it.

So Where Does That Leave Me?

I wrote a whitepaper about agentic commerce as a competitive advantage. I still believe that’s true.

But I also can’t unsee the pattern now.

Every industry that went through this transition experienced the same arc: efficiency gains first, market structure problems second, regulation third. Finance got the SEC after the crashes. Advertising is still fighting fraud. Energy exposed regular people to volatility without protection.

Retail is entering the efficiency phase. We’re celebrating the 7x growth differential. We’re talking about conversion optimization and attribution models.

We’re not talking about what happens when the algorithms misbehave. Or when bad actors figure out how to game the demand signals. Or when consumers accidentally expose themselves to price volatility they don’t understand.

Maybe we’ll figure it out before the crash.

Or maybe—like every other industry—we’ll build the guardrails after.

The whitepaper describes what’s coming. This post is me realizing we’ve seen this movie before—just in different theaters.

The question isn’t whether retail will follow the same arc. It’s whether we’ll learn anything from the industries that already walked this path.

For now, I’m watching. And taking notes.