Maybe browser tabs will become a source of eCommerce nostalgia. Because consumers who once toggled among a dozen browser tabs and comparison-shopping sites are increasingly posing a single question to an LLM such as Perplexity — and pressing “buy” on the answer it returns.
It’s agentic AI at its most promising. That subtle shift in search behavior is forcing merchants to rethink where, and how, they meet shoppers. If an AI agent can compress browsing, selection and checkout into the same dialogue, retailers that sit outside the conversation risk ceding both visibility and sales.
Behind the scenes of that are the software agents empowered not merely to fetch information but to take actions on the customer’s behalf, from adding an item to cart to completing payment. The technology marries an LLM’s knack for understanding natural-language requests with direct hooks into merchant catalogs and payment rails. “What consumers really want is for commerce to happen immediately,” Scott Hendrickson, chief revenue officer of the agentic AI merchant network firmly, said in an interview with PYMNTS CEO Karen Webster. “People’s time has never been more valuable, and they expect things to be frictionless.”
Hendrickson and his co-founder Kumar Senthil argue that the evolution from search to suggestion to settlement is the next logical step in eCommerce. Social media feeds already entice users with one-click purchases of impulse items they “didn’t know they wanted,” Hendrickson noted, pointing to TikTok Shop’s rise as evidence that contextually embedded commerce can move high volumes of goods. Agentic AI extends that concept to any venue where a shopper starts research, whether it’s an email thread, a mapping app, even a voice assistant in the kitchen. Wherever the conversation starts, the agent can surface suitable products, verify availability and finalize the order in-line.
Why the Existing Funnel Breaks Down
Today’s hand-off from Google or an affiliate blog to a merchant checkout page is riddled with what Senthil calls “banana peels.” High-intent shoppers often abandon their carts when they are asked to create yet another account, re-enter card credentials or confirm whether a promised promotion code applies. “You have to hit them at that moment, otherwise the customer walks away and you lose,” Senthil said. The longer the distance between discovery and payment, the more likely a sale is to slip away.
Other frictions lurk deeper in the stack. Browser automation — the technique many early chatbots use to mimic a human clicking through webpages — stumbles when confronted with a shipping address a retailer won’t accept or when inventory data are a few minutes stale. The result can be a botched order that reflects poorly on both merchant and AI provider.
Firmly’s answer is a real-time integration hub that plugs directly into merchants’ existing application-programming interfaces, or APIs. Rather than scrape public product pages, the network ingests live catalog data, inventory counts and price changes, then exposes that information to partner platforms such as Perplexity in what Senthil said adds “about 50 milliseconds” of processing time. If the agent selects a red toaster, the system already knows whether the model is in stock, whether the retailer ships to apartments, and what forms of payment it accepts. “If the inventory is there and the price is right, the order has to get placed,” Senthil said, adding that direct API calls also preserve audit trails that major brands need for compliance.
To keep latency down, firmly pre-caches frequently requested data and streams order confirmations back to the AI in real time. That architecture, Senthil said, is critical if conversational commerce is to feel as instantaneous as consumers have come to expect from one-click checkout buttons elsewhere on the web. A lag of even a few hundred milliseconds risks reminding users that a complex transaction is unfolding behind the curtain.
Consolidating the Purchase Funnel
For Hendrickson, the technical plumbing enables a strategic shift in how merchants think about customer acquisition.
“Right now, the search experience is you search for a product and then you immediately leave,” he told Webster. “With the new LLMs you might start your research, refine what you’re looking for, and make the purchase in the same place. You don’t have to leave the site to complete that full funnel anymore, and that’s what’s ultimately going to drive better performance.”
Consolidating the funnel could also upend traditional advertising economics. If the agent closes a sale within the first interaction, retailers may decide to shift dollars from pay-per-click campaigns toward deeper catalog and logistics integrations. Conversely, platforms that mediate the sale, such as chatbots, voice assistants or augmented-reality overlays, gain new leverage to capture a slice of the transaction rather than bill for impressions.
That model works only if the agent can reliably see a broad, up-to-date range of products. Firmly says its network already covers about 11 million merchants, representing a majority of U.S. online gross merchandise value. By federating those integrations, the company hopes to spare each newcomer to conversational commerce the months-long process of onboarding retailers one at a time.
“The bulk of the $1.3 trillion in U.S. eCommerce GMV is now available for these platforms to use,” Hendrickson said. Relying on a single platform API such as Shopify’s, he added, would expose only a slice of the market and leave out many big-box chains and specialty retailers that run bespoke systems.
What Merchants Want—and Fear
The prospect of selling through a disembodied agent raises as many questions as it answers. According to Senthil, merchants’ first reaction is often: Who owns the data and the customer relationship? Firmly’s response is that the retailer remains the merchant of record, preserving liability protections, access to order information and the ability to offer loyalty rewards. Integration, he said, generally requires little more than whitelisting an IP address and sharing API credentials. “It’s music to their ears” for resource-constrained operators juggling tariff uncertainty, rising advertising costs and the shift to omnichannel fulfillment.
As Webster pointed out, some brands could balk at the idea that a sale could occur without a click-through to their carefully curated storefront. Lifestyle labels in particular worry that ceding the front-end experience to an agentic AI could erode brand storytelling. Hendrickson counters that many customers now encounter products through a third-party channel anyway from an influencer’s livestream, a buy button inside Instagram, or a same-day-delivery marketplace.
“We’re not eliminating brand,” he said. “We’re shortening the distance between intent and conversion.”
If post-pandemic consumers taught merchants anything, it is that convenience quickly becomes table stakes. Three years ago, curbside pickup and one-click checkout were differentiators. Today, shoppers abandon carts when those options are missing. Senthil predicts the same acceleration curve for conversational commerce. Google has estimated that only about 7% of product searches result in an immediate transaction. Agentic AI, he argues, can recapture a chunk of the remaining 93% by answering detailed questions: Will this jacket fit someone 6 feet tall? Does it ship before Friday? Then it can execute the order before doubt creeps in.
“People with high intent are just going to switch to this model,” he said. “They want the research time to drop from hours to minutes.”
Early use cases are telling. Shoppers ask a voice agent to “reorder my favorite espresso capsules” or “find a mid-century lamp under $150 that fits on a 24-inch table.” The agentic AI responds with a shortlist, factors in personal preferences and loyalty programs, and, with permission, places the order. Over time, the same workflow could handle more considered purchases: a premium kitchen appliance that requires comparing wattage, finish and warranty terms, or a flight-hotel-tour package that bundles several suppliers.
The shift won’t happen overnight, and skeptics point to the patchy track record of chatbots to date. But Hendrickson argued that incremental experiments today will pay off when agentic experiences hit the mainstream.
“Merchants have to lean in,” he said. “This world is evolving quickly, and we can test — and, most important, learn — without a lot of cost. Those who adopt early will see the benefits, just as they did in the early days of social commerce. It’s a chance to learn. Don’t be afraid of it.”
