AI Commerce News Digest (August 19, 2026)
A Shopify merchant built a production AI commerce agent in 10 weeks, OpenAI scaled back Instant Checkout, and 11st Street doubled conversions with generative AI search. Here is the AI commerce news for August 19, 2026.
Key Takeaways
- A Shopify merchant built a production AI commerce agent in 10 weeks
- OpenAI scaled back Instant Checkout and refocused on discovery
- Payment networks and retailers are converging on the discovery layer
Here is the AI commerce news for August 19, 2026. Today the idea that AI agents will handle the buying itself ran up against implementation and evidence. Concrete builds moved forward, including a merchant's in-house agent and a retailer wiring its store network into agent-driven demand, while reports lined up showing that in-chat checkout is not seeing the use its backers expected. In contrast with Alipay's full-stack infrastructure push covered yesterday, attention is shifting toward the accuracy of product discovery.
Today's Top Stories
Redmond built a production AI commerce agent in 10 weeks on Shopify's Storefront MCP
Redmond, mining natural salt in Utah since 1958, built a production AI commerce agent in 10 weeks on top of Shopify's Storefront MCP.
www.shopify.comShopify published a case study on Redmond, a Shopify Plus merchant that built a production AI commerce agent in 10 weeks on top of Storefront MCP. Redmond has mined natural salt from an ancient deposit in central Utah since 1958 and now runs more than 10 brands, including Real Salt and the Re-Lyte electrolyte line.
The trigger was the discontinuation of the managed AI customer service tool the company had been using. Pay another vendor, or build. Redmond chose to build, with a two-person team: Phillip Hinson, an applied AI developer, and Jeremiah Payne, the development team lead who had managed the company's Shopify infrastructure since 2018. The agent went live in February 2026 and is already handling customer conversations at scale.
What stands out is not the adoption of another SaaS product but that building on a protocol layer turned out to be feasible within a realistic effort budget. Storefront MCP exposes catalog, cart, and order data in a form agents can work with, so a merchant keeps its own inventory and pricing while designing only the conversational layer. For anyone weighing a replacement for a customer service tool, "build it" has entered the comparison set.
Full article: Two People, 10 Weeks: Inside the AI Commerce Agent Redmond Built on Shopify's Storefront MCP
OpenAI scaled back Instant Checkout as trust in AI payments lags
The Economic Times traces why OpenAI narrowed its Instant Checkout ambitions and shifted priority to product search and discovery, with reactions from Indian ecommerce operators.
m.economictimes.comThe Information reported that OpenAI has scaled back Instant Checkout and shifted its priority to product search and discovery, and The Economic Times covered the story alongside reactions from India's ecommerce market. Launched in September 2025, the feature let users buy items from Etsy, Walmart, and Shopify inside ChatGPT, but adoption did not follow.
The numbers are pointed. In the Walmart experiment, conversion through ChatGPT checkout ran about a third of traditional channels, and the retailer refocused on its own shopping assistant, Sparky. Discovery, meanwhile, is performing well: Flash AI reports that traffic arriving through AI surfaces converts at more than three times the rate of non-AI traffic. Morgan Stanley projects AI shopping assistants could drive up to $385 billion in US ecommerce spending by 2030.
Retailer caution also has a data dimension. With no clarity on how much first-party customer data, including purchase records and contact details, will remain accessible, handing checkout to a chat interface is a hard call to make. For now the contest is over discovery rather than payment, and that view is being shared with measured numbers behind it.
Full article: AI Traffic Converts 3x Better, Yet Agentic Checkout Is Still Stalling
Agentic Commerce
JB Hi-Fi wires natural-language search and agent buying into its store network
Australian consumer electronics retailer JB Hi-Fi told investors it will introduce natural-language product search and agent-based shopping and connect them to its stores and marketplace.
www.mi-3.com.auJB Hi-Fi, one of Australia's largest consumer electronics retailers, used its results announcement to tell investors it will introduce natural-language product search and agent-based shopping experiences and connect them to its store network and membership programs. An expanded marketplace and the integration of phone and chat service channels are part of the same plan.
Record trading sits behind the move. Revenue reached A$11.06 billion, leaving the balance sheet with room to invest. Capital expenditure rose 6.4% to A$87.4 million. In-store retail media screens will expand from about 140 to 250, and electronic shelf labels will reach 100 Australian stores, with management expecting ticketing work to give way to customer service.
The ground is not level, though. Comparable sales in Australia are weakening, and shoppers increasingly hold out for major promotional events. An ACCC action added roughly A$14 million in cost. How a retailer with physical stores connects its traffic and staff to agent-driven demand is the open question, and putting that allocation on record this early is what makes the disclosure notable.
Alipay's Abao agent reaches cars and phones, with the AHA protocol released
Alipay connected its Abao super agent to five phone brands and sixteen carmakers and released AHA, a cross-device protocol for interconnecting agents.
pandaily.comAt Alipay's first AI ecosystem conference on August 17, the company revealed that its super agent Abao now connects to five phone brands and the in-car systems of sixteen carmakers. It follows the full-stack infrastructure announcement covered yesterday, and merchant demand for integration is reported to be queued into next year.
Alongside it came AHA (Agent Hub Access), a protocol for interconnecting agents across devices. The design addresses integration effort and coordination safety through domain-level authorization and data isolation. Through Agentar from Ant Digital, merchants convert mini-programs into standard Skills that distribute to phones, car infotainment, smart glasses, and watches in one pass. More than 200 merchant Skills are already registered.
At the venue, reporters ordered KFC by voice from a car, and McDonald's, Luckin Coffee, Amap, and Didi joined alongside logistics operators Deppon, YTO, and Hive Box. Ant Group CEO Han Xinyi said super agents will spread within six to twelve months. Points of contact are dispersing toward the device layer, which is reason enough for merchants elsewhere to revisit how their product data is structured.
AI Commerce Tools
11st Street's generative AI search more than doubled conversions and went to full rollout
South Korean marketplace 11st Street recorded purchase conversion rates more than double standard keyword search in its generative AI search pilot and rolled it out across major categories.
www.techtimes.comSouth Korean marketplace 11st Street announced that the generative AI search built into its mobile app produced purchase conversion rates more than double those of standard keyword search, and on August 18 it moved to a full rollout across all major categories. The feature had been deployed quietly the previous month.
The design hinges on not returning a flood of results the moment a query arrives. Type "laptop" and the AI interprets the category, surfaces granular sub-intent keywords such as "lightweight and portable" or "optimized for gaming," and waits for the shopper to choose before retrieval begins. A second AI layer then applies multi-criteria analysis across price competitiveness, delivery cost, delivery speed, and review scores, narrowing to roughly five recommendations.
A natural-language mode called situational search handles phrases like "gifts for a housewarming" without keyword entry. Settling intent before ranking results is an ordering change that also matters when an AI agent is the one choosing products. It is a useful prompt to check whether your own product data can withstand filtering by criteria rather than by keyword.
Full article: 11st Street's AI Search Doubled Purchase Conversion: What Narrow-Then-Search Means for E-Commerce
Malachyte brings Spotify-style behavioral reading to ecommerce with $10 million raised
The team behind Spotify's recommendation infrastructure is applying behavioral intent reading to ecommerce, backed by a $10 million round.
www.pymnts.comMalachyte, founded by members of the team that built Spotify's behavioral intelligence infrastructure, raised $10 million to apply that work to agentic commerce. Co-founder and CEO Sidd Motwani told PYMNTS the infrastructure powers more than 90% of Spotify's recommendations across 800 million users and a billion items.
The core idea is how to read a skip. Skipping a track does not necessarily mean dislike: a song that misses during a morning run can be exactly right on the drive home. Behavior carries meaning only inside the context that produced it. Retail works the same way, and Malachyte reads scrolling past an item, dwelling on another, checking a size, changing a price filter, backing out, and adding something else as a sentence rather than a set of words.
The company says it can read the current situation within 200 milliseconds even for visitors with no history. Merchants retain sliders over inventory, margin, and promotional weighting, which positions the product as an orchestration layer rather than a recommendation widget. A caveat travels with it: systems that take every signal literally tend toward showing shoppers more of what they have already seen, lifting conversion on paper while thinning discovery.
Bridgeline's HawkSearch deployed across 2.4 million automotive product records
Bridgeline Digital deployed HawkSearch for a major North American automotive ecommerce network, covering about 2.4 million product records.
www.stocktitan.netBridgeline Digital announced that it deployed HawkSearch for a major North American automotive ecommerce network covering about 2.4 million product records, one of the largest OpenSearch implementations the platform has run.
The gains show up in operations. Product indexing time dropped from roughly 20 hours to 3 hours, and hierarchy facet response times improved by about 91% to around 175 milliseconds, with real-time indexing enabled. The deployed capabilities include smart search, visual and concept search, AI-assisted merchandising, and personalized recommendations.
This is phase one of a multi-phase modernization. Thirteen B2C sites migrate next, followed by B2B platforms and a planned hybrid search implementation. For catalogs where part numbers and heavy filtering are unavoidable, search response time translates directly into abandonment.
Payments and Fintech
Rain launches the Agentic Payments Alliance with Visa and Mastercard as founding members
Stablecoin payments infrastructure company Rain formed the Agentic Payments Alliance to work through how AI agents are authorized and how fraud is detected.
www.pymnts.comStablecoin payments infrastructure company Rain announced the Agentic Payments Alliance (APA) on August 18, with Visa and Mastercard among its founding members. Fiserv, Circle, Solana, and Remitly have also joined.
The agenda covers what has not been settled: how AI agents are authorized, how fraud is detected, and how loyalty and rewards will work. Rain co-founder and CEO Farooq Malik said no single company should decide how agents transact on someone's behalf, and described the intent as putting platform builders, regulators, and the developers closest to actual agent usage in the same room. Rain's existing offerings include its Agent Control Layer and Scoped Cards.
The announcement cites McKinsey projections of $3 trillion to $5 trillion in global agentic AI commerce by 2030. Coalitions of this kind tend to lead with statements of direction, though, and actual specifications and operating rules take time to appear. Several similar bodies have formed this month, and which of them reaches implementation is not yet clear.
Consumer Trends
Capgemini sets out three requirements for agentic CX, with 58% of consumers expecting time savings
Capgemini lays out three areas retailers should address as customers begin relying on AI agents across discovery, purchase, and support.
www.capgemini.comCapgemini published a report organizing what retailers should address as customers begin relying on AI agents. Research from its institute found that 58% of consumers believe AI agents can save them time on routine monthly purchases and payments, and 68% of organizations expect AI agents to outperform traditional CX channels.
Three requirements are set out. First, organize experiences around customer intent rather than channels or departments. Shopping ahead of a trip and replacing a failed product call for different things even when both begin with similar queries. Second, connect agents to the internal systems that let them act. An agent with no reach into inventory, pricing, promotions, or fulfillment status can only return information.
Third, preserve context when a person takes over. If the employee cannot see what the agent proposed and what it already did, the customer ends up repeating themselves. The framing assumes that specialist judgment and empathy remain necessary, which makes it a useful structure for revisiting service design.
72% of retail apps are downloaded and deleted, Adobe finds
An Adobe survey of 1,000 US consumers found 72% had downloaded a retail app for a one-time discount or purchase and eventually deleted it.
www.ecommercetimes.comAn Adobe survey of 1,000 US consumers found that 72% had downloaded a retail app for a one-time discount or purchase and eventually deleted it. Installs rise while retention does not follow.
Reasons for installing were joining a loyalty or membership program at 45%, a one-time discount or promotion at 42%, and an easier shopping experience at 37%. Reasons for uninstalling were no longer needed for the original purpose at 64%, too much storage at 43%, and lack of ongoing relevance at 43%. The notifications that drive engagement are practical ones: order status updates at 39%, personalized discounts at 37%, and price drops on saved items at 33%.
A partner at Kearney observed that most retail apps have an acquisition strategy but not a retention strategy. If the app is a smaller version of the website, there is no reason to give it space on a phone. One in three respondents said they would be more likely to keep an app that offered voice or image search, suggesting that AI-native input can itself become a reason to stay.
Global Ecommerce Trends
US ecommerce share reaches 17.1%, with quarterly online sales at $340.2 billion
US Census Bureau data shows ecommerce reached 17.1% of retail sales, with seasonally adjusted online sales of $340.2 billion.
www.pymnts.comUS Census Bureau quarterly data shows that ecommerce accounted for 17.1% of retail sales in the second quarter of 2026, up from 16.3% a year earlier. Seasonally adjusted online sales reached $340.2 billion, up 3.8% from the prior quarter and 12.2% year over year. Total retail sales rose 6.7% year over year.
The share is climbing for a simple reason: online is growing faster than retail overall. Research from PYMNTS Intelligence and Visa Acceptance Solutions found the typical American shopper now records 51 digital shopping days a month, up from 46 in January 2024. Digitally assisted shopping inside stores rose from six days to seven.
The same study found that 60% of shoppers want price matching while only 47% of merchants provide it, the largest gap between demand and availability it measured. Online, switching stores is easy, so how price presents itself feeds directly into abandonment.
Wayfair's Q2 US growth hits a post-pandemic high on AI and Perigold
Wayfair reported its strongest US growth since the pandemic in Q2 2026 and credited AI capabilities and its luxury brand Perigold.
www.marketscale.comHome goods retailer Wayfair posted its strongest US revenue growth since the pandemic in the second quarter of 2026, crediting AI capabilities and its luxury brand Perigold. Notably, the company framed AI tooling as a current revenue driver rather than a long-term investment with a delayed payoff.
Perigold serves a higher-value buyer with curated premium selection, separate from the core Wayfair storefront. That the segmentation is contributing distinctly says something useful about how large digital retailers structure brand architecture.
In the same period, Kohl's launched a Google Gemini-powered shopping assistant ahead of the back-to-school season. A national chain putting it in front of customers during a peak conversion window, rather than in a pilot, makes it a more useful benchmark for vendor selection.
Corporate Moves and Partnerships
Prodx and Hy-Vee extend their partnership across 265,000 products
Product data infrastructure company Prodx extended its partnership with Midwest grocer Hy-Vee, covering roughly 265,000 products.
www.prnewswire.comProduct data infrastructure company Prodx announced an extension of its partnership with Midwest supermarket chain Hy-Vee. Prodx has structured nearly 265,000 Hy-Vee products, standardizing an average of seven attributes per SKU across more than 1,200 attribute types and 5,800 active categories.
That foundation supports buy-again, substitutions, personalization, and digital merchandising. Returning shoppers see Purchases and Essentials experiences that surface frequently bought items in the order they usually shop, while Predictions anticipates needs before a search happens. The Prodx Aisles App gives Hy-Vee direct control over merchandising, A/B testing, and contextual relevance.
Hy-Vee operates more than 560 business units across nine Midwestern states with annual sales above $14 billion. The more attention agent-driven discovery attracts, the more this unglamorous groundwork matters. The accuracy of substitutions and repeat purchases is decided by how granular the attributes are.
Closing Thoughts
The idea of handing over payment paused for a moment today, and investment leaned toward discovery instead. ChatGPT's checkout narrowing, with Walmart conversion running at about a third of traditional channels, draws a concrete line under what agents can currently be trusted to do.
Implementation kept moving all the same. A merchant built an agent in-house in 10 weeks, a Korean marketplace reordered its search flow and doubled conversion, and an Australian retailer moved to connect its entire store network to agent-driven demand. The common thread is that the work is happening beneath the conversational surface, in product data, inventory, and the interpretation of intent.
From here, watch whether frameworks like the Agentic Payments Alliance turn into actual specifications, and whether Alipay's AHA protocol travels beyond China. Until payment standards settle, the preparation available to merchants stays concentrated on getting product data in order.