Loops Raises $3.5M From a16z Speedrun to Build the Retail OS for Agentic Commerce

Seven out of every ten shopping carts filled online are abandoned before anyone pays, according to the Baymard Institute, which puts the global average at 70.22% after combining 50 separate studies. That number says a lot about where ecommerce stands today, because online stores have become brilliant at showing people thousands of products while remaining surprisingly bad at helping them settle on one. Loops, a San Francisco startup that builds AI for brands and retailers, has now raised a $3.5 million seed round to close that gap. The round was led by a16z speedrun, the startup accelerator inside Andreessen Horowitz, with NTT Docomo, Connecticut Innovations, Alumni Ventures and Monday Capital also taking part. https://www.youtube.com/watch?v=VPSkUJFD_-U&feature=youtu.be&embedable=true Rather than selling retailers one more AI tool, Loops is building what it calls an operating system for retail, a single AI layer that runs search, shopping assistance, customer support, product data clean-up and virtual try-on, five jobs that most retailers currently buy from five different vendors. Because all five share the same understanding of the catalog and of what each shopper is trying to find, every part of the system gets sharper as the others learn. The company will use the new capital to build out its product, add capacity for deploying with customers and expand into new countries.Agentic Commerce Is Measured in Trillions Agentic commerce describes shopping in which an AI agent does part of the work on the buyer's behalf, searching the catalog, comparing products, answering questions and in some cases completing the purchase itself. The market it is growing into is already enormous, with worldwide online retail sales reaching $5.1 trillion in 2022 according to eMarketer figures published by Shopify, on the way to an expected $6.9 trillion in 2026 and $7.9 trillion by 2028. McKinsey expects AI agents to orchestrate about $1 trillion of US retail sales by 2030 and between $3 trillion and $5 trillion worldwide, which means that at the top of its range, agent-led shopping alone would match everything the world bought online in 2022. For retailers, the practical consequence is that a fast-growing share of their customers will begin the journey with a conversation rather than a search box, so the stores that can hold that conversation well are the ones that will win the sale.Where Online Stores Lose Their Buyers Online stores start losing buyers long before anyone reaches checkout, with failed search doing much of the damage. A Harris Poll survey of 13,500 shoppers for Google Cloud found that 53% leave a retailer's site when search fails to turn up even one item they want. Google estimates that failed search costs retailers more than $2 trillion a year worldwide, including $234 billion in the US alone, while the reverse is just as telling, since 78% of US shoppers end up buying at least one extra item after a search that actually works. Returns are the second big leak for online stores, with the National Retail Federation expecting Americans to send back $849.9 billion worth of goods in 2025, equal to 19.3% of everything bought online. A large share of those returns comes down to the simple fact that a shopper could not judge size, fit or look from a flat product photo, which is exactly the kind of problem that richer product information and virtual try-on are built to solve.AI-Referred Shoppers Now Buy More Shoppers who arrive at a store from AI tools such as ChatGPT have gone from being the weakest visitors a retailer had to being the most valuable in the space of a single year. In March 2025 they converted 38% worse than everyone else, according to Adobe data reported by TechCrunch, yet by March 2026 they were converting 42% better than other visitors and spending 37% more on every visit. The volume is rising just as quickly, with AI traffic to US retail sites up 693% over the 2025 holiday season and 393% in the first quarter of 2026, while about 39% of US adults now say they have used AI to shop online. These shoppers are used to describing what they want in plain words and getting a straight answer back, which is not something a store built on keyword search and filter menus was ever designed to give them.The Weak Link Is Product Data An AI agent can only recommend a product it can properly read, which is a real problem for retailers because Adobe's own checker found that product pages are the least readable part of a retail site, with only 66% of their content visible to AI systems. Returns, contact and FAQ pages score around 80% on the same test, which means most retailers have made their help pages easier for AI to understand than the very pages that are supposed to sell. That gap matters because product pages hold the facts a shopper needs before deciding, so when details such as size, material or fit are missing, both the store's own search and outside AI assistants end up giving weak answers. Loops treats product data as the base layer of its retail OS, using its enrichment tool to rewrite descriptions and fill in missing attributes so that search and the shopping assistant always work from accurate, complete information.The Founder Thesis Loops began with a lost sale in Bali, where co-founder İlker Selim Zorluoğlu tried to buy a gift for his two-year-old daughter, messaged the retailer with a question and then waited hours for a reply that came too late, so the gift was never bought. When he and co-founder Ari Nazir, now the chief executive, dug into why that happened, they found a much bigger gap than slow customer support. Shoppers were already telling AI exactly what they wanted in plain language, yet most online stores still ran on keywords, filters and separate support queues. The first prototype, which Nazir built over a single weekend, has since grown into a company with more than 70 customers and pilots that has handled millions of shopper interactions. The idea behind it is that retailers should stop buying AI one feature at a time, because today one tool improves search, another answers questions and a third writes product descriptions without any of them sharing what it learns with the others.Zorluoğlu, who serves as president, says the biggest mistake he sees is treating AI as a collection of features, arguing that a retailer really needs one system that understands the catalog, the business and what the shopper is trying to do. Inside Loops each part feeds the next, so better catalog data improves search, search gives the assistant context and customer conversations reveal where product information is still weak. Virtual try-on then lets shoppers see an item on themselves before buying, which tackles returns at the source. Early Proof at Retail Scale Turkish discount supermarket chain A101 uses Loops to help customers find products and get answers throughout the purchase. Sessions involving the platform showed roughly 83% higher purchase conversion than a comparison group, a result the company is careful to describe as observational rather than a controlled test. Denim brand Mavi uses Loops for AI search and shopping help, which lets its customers ask for jeans by style, occasion or fit in much the same way they would ask a sales assistant in a store. For a seed-stage company, the most telling number is a commercial one, since 64% of Loops pilots have already turned into paying customers. Converting nearly two in three trials into signed contracts this early is a strong sign that retailers see the value quickly and are willing to pay for it.The Investor Thesis Fareed Mosavat, investment partner at a16z speedrun, says AI is changing the way companies and their customers interact, though he argues that the platforms that last will be the ones that work reliably inside real businesses. Commerce stands out to him because the link between a better AI experience and higher revenue is unusually clear, which is why he points to Loops' pilot conversion and early customer expansion as evidence that brands want one platform able to own more of the shopping journey. The money already moving through AI apps supports that view, with eMarketer forecasting that US shoppers will spend $20.6 billion through AI platforms in 2026, nearly four times the 2025 level, before that figure passes $144 billion by 2029. That sevenfold rise in three years will flow only to stores whose products an AI system can find, read and explain to a buyer, which is precisely the layer Loops is building. The market has also shown how much a single piece of that journey can be worth, since site-search specialist Algolia was valued at $2.25 billion in 2021 while discovery platform Bloomreach reached $2.2 billion in 2022, both by doing search and discovery well. Loops runs search as just one of five jobs on a single retail OS, so each new tool a retailer switches on makes the others more useful because all of them learn from the same catalog and the same shoppers.Alongside a16z speedrun, which invests up to $1 million in each company it selects for its cohort, the round brings in NTT Docomo, Japan's largest mobile operator, as a strategic backer with long experience in consumer technology across Asia. Connecticut Innovations, Alumni Ventures and Monday Capital complete the seed, giving Loops a mix of venture funds and a corporate investor as it prepares to take its retail OS into new countries.From Retail to Other Industries Loops sees commerce as its first market rather than its only one, because it believes the lasting value in enterprise AI will sit with systems built deep inside a single industry. Those systems carry the knowledge, integrations and controls needed to run real work, which a general model with a thin layer on top rarely can. Financial services and energy are next on its list, with the company also building small language models for tasks where a focused model fits better than a large general one. The playbook stays the same in every new industry, which is to enter through one clear problem with obvious value, learn how that industry actually works by staying close to customers and then turn what repeats into product. Retail is where that approach is being proven first, at the very moment agentic commerce is moving out of consultants' forecasts and into everyday shopping.Don’t forget to like and share the story! Vested Interest Disclosure: HackerNoon has reviewed the report for quality, but the claims herein belong to the author. #DYOR.