Designing the Checkout That Survives Agentic Commerce

Retailers have spent years perfecting the top of the funnel. Ad targeting gets sharper every quarter, product pages are A/B tested to the pixel, and personalization engines surface the right item at the right moment. Then the shopper reaches checkout and runs into a forgotten password, a sluggish wallet load, or a five-step login flow that didn't need to be five steps. That gap is where revenue disappears, and most retail organizations have never fully reckoned with the cost because they still treat payment infrastructure as a back-end concern.

Dhvani Unadkat, a product marketing manager, has built her career around closing exactly this kind of gap. She currently leads product growth at PayPal, and was a product marketing manager at Amazon prior to that where she worked in product marketing on AI-powered audience and targeting tools for advertisers. Her decade across ad tech, product marketing, and analytics, paired with a recent Harvard Business School Online certificate in AI for Business, has given her a working theory of where intent goes to die.

"Checkout is not the last part. It's where you make or break the customer's journey," Unadkat says.

The deliberation window

Tech production cycles in retail are slow by nature, and by the time an update ships, consumer expectations have moved again. That velocity gap leaves specific pressure points unaddressed, and each one becomes a place where intent dissolves into hesitation.

"By the time you bring it up to speed, it's already changed and the customer wants something else," Unadkat says. A forgotten password, a card the system doesn't recognize, a wallet that takes two seconds longer than expected, any one of them creates room for the shopper to reconsider and for the purchase not to happen.

Time equals doubt. The longer a checkout takes, the more likely a shopper is to second-guess the decision, and retail runs on impulse more than almost any other category. The deliberation window is what kills it, which is why the conversation across the industry has shifted toward collapsing that window through biometrics and one-tap flows.

"Working on the human psychology of being in the moment and making that instinctive decision, coupled with biometrics, is what will lead to higher conversion numbers," Unadkat says. "If I can complete a payment fast, I have less time to think about whether I'm actually thinking through what I'm buying or making an impulsive purchase."

Biometric login collapses the window almost entirely. Face ID, fingerprint, done. Attention at Shoptalk and other recent industry events reflected exactly that interest, with a wave of vendor activity around connecting fast biometric authentication to streamlined payment flows. The checkout itself becomes nearly invisible, and a checkout no one notices is a checkout no one abandons.

The Uniqlo benchmark

The reference point Unadkat keeps returning to is a trip to Japan, where she used Uniqlo's RFID-enabled checkout. Shoppers drop a basket of clothes into a scanning bin, see an instant total, pay, and leave.

"Whenever I need something, my first go-to is Uniqlo, because I know I can get everything done fast and won't spend a lot of time pondering," she says. "When you're in those lines, you're thinking about whether you need this or that. You end up removing three or four things from your cart."

Speed becomes a loyalty driver, and shoppers who trust an experience to be fast keep returning to it for reasons that have nothing to do with a rewards program. The experience earns the repeat visit that points and tiers are usually built to buy.

The same logic translates directly to digital, and Unadkat points to Apple Pay. "Regardless of what you buy, the experience for Apple Pay is two clicks, it's a biometric login, and it's always a half-sheet that comes up to complete your payment," she says. "You're not trying to figure out if you've done anything right or wrong. It's just always the same."

Predictability at checkout outperforms novelty every time, and most retailers have the order reversed.

The personalization blind spot

There is a data problem hiding inside checkout that most retailers have not addressed. Payment systems can identify returning customers easily enough, but they cannot yet understand context, and context is where the real signal lives. The same shopper makes different payment decisions depending on where she is buying and what is in her cart.

A frequent shopper might reliably use one wallet for beauty purchases and a different one for electronics, sorting by order value or by category-specific trust without ever thinking about it. No platform currently aggregates that behavioral signal and gives retailers a propensity score for payment method preference by context, which means a major personalization lever is sitting unused.

"Is there an aggregator that can tell you that one customer, when she shops with Sephora, she's going to use Apple Pay, versus when she's shopping with Best Buy, she's going to use PayPal because of the average order value or the kind of purchase she wants to make?" Unadkat asks.

That signal, surfaced at the right moment, changes what gets offered and when, and it extends personalization into a layer most brands have not built. This is the kind of contextual intelligence that separates real 1:1 strategies from segmentation that just sorts shoppers into bigger buckets.

Checkout as trust infrastructure

Agentic commerce raises the stakes, and it is arriving faster than most payment platforms can respond to. Several large merchants are already testing AI shopping agents that browse catalogs and assist with decisions, and whether those agents can ever complete payments depends on whether the same friction points that lose human shoppers also stop a bot. If each merchant builds a different agent with a different authentication flow and a different payment method, the fragmentation that already plagues digital wallets gets baked into the AI layer permanently. An agent that hits a five-click login wall stops there.

Shoppers themselves are not consistent, and that is the harder problem to solve. Reordering household staples, researching a home decor purchase, and buying a high-ticket gadget involve different decision architectures, and no single agent handles all of them the same way.

"There are customers like myself who would really like to go to the Macy's app and look through all the options, versus somebody like my husband," Unadkat says. "He would hate it and show me a black T-shirt and send it to my house. Different retailers, depending on what they're trying to shop for, their shopping habits change and the agentic experience has to adapt."

Most of the agentic activity Unadkat encounters today is still in assistance mode rather than transaction mode, with agents helping shoppers browse rather than completing the purchase themselves. The infrastructure question is already live, though, and the consumer trust gap only widens for brands that have not standardized fast, consistent, low-latency checkout. Those brands will be deprioritized by agents before a purchase decision is ever made.

The implication for marketers runs deeper than the checkout screen itself. Owned channels, the SMS list, the email file, the push subscribers, become the last direct line to a customer once agents are running discovery on the shopper's behalf, and a checkout experience that breaks the trust built through those channels undoes everything the upstream work was meant to accomplish.

"Every payment you make is adding to the bottom line of the organization that is selling that," Unadkat says. "It's the heart of the entire process for the retailer and the customer. And providing that ease is actually a way to the heart of a customer wanting to shop with you."

The brands that win the agentic era will be the ones whose checkout disappears entirely, and the work of making it disappear starts now.

The views and opinions expressed are those of Dhvani Unadkat and do not represent the official policy or position of any organization.