Inside AI’s $5 trillion quest to develop taste

FastCompany: The assignment is charming: I’ve been asked to moderate a panel in a garden in early summer. The problem is that it will require an outfit I don’t own and have no time to find. Easy enough, I think. I use AI every day to summarize transcripts, synthesize financial data, and draft emails. Surely I can outsource this. I open ChatGPT and ask it to find me a dress from my favorite label, Sézane, that’s appropriate for an outdoor professional event. It pulls dresses from the French brand’s catalog that are made of linen and crocheted—floaty, unstructured things that require a level of undergarment coordination I don’t want to think about.

I clarify. It pivots. After several rounds of negotiation, it finally surfaces a red polka-dot dress I like. There are no links. I go to Google, search manually, and discover that the dress is from last season and available only in a size 2 on Poshmark. Twenty minutes after starting my search with ChatGPT, I give up. I head over to Sézane’s website and buy something the old-fashioned way. Artificial intelligence has, in the span of a few years, displaced coders, passed the bar exam, and written term papers for an entire cohort of college students. Yet, the technology that’s remaking civilization still can’t help me buy a dress.

The retailers who crack this code are likely to own the next era of online shopping. When ChatGPT launched in 2022, becoming the fastest-adopted consumer technology in history, it lit a fuse. Behemoths like Amazon and Walmart, along with smaller players, raced to roll out AI-powered assistants to help customers find the right product from their catalogs. Soon startups like Daydream, Phia, and Remark were pulling in millions in venture capital funding to build AI shopping agents that could offer hyperpersonalized product recommendations.

Over the past year, OpenAI, Google, and Anthropic have been working behind the scenes with the e-commerce infrastructure giants Shopify and Stripe to build essential retail plumbing , such as AI-native checkout systems and real-time inventory updates. They’re preparing for a world of agentic commerce , where autonomous agents browse, compare, and buy on your behalf. Very soon, these companies promise, your AI assistant will know you well enough to anticipate your weekly grocery run and order the gear you’ll need for an upcoming ski trip, without you lifting a finger. Already 2% of all ChatGPT queries—50 million a day—involve shopping.

McKinsey estimated last October that AI assistants could enable up to $1 trillion in U.S. shopping annually by 2030, and up to $5 trillion globally . But as with so much AI-enabled technology, there’s a chasm between promise and reality. Right now, the experience of AI-assisted shopping is frustratingly ineffective. Walmart’s Sparky, perhaps the most advanced shopping chatbot, is still only slightly better than a search bar when it comes to certain things. Most frontier models, meanwhile, can’t tell if a product is in stock, much less let you click to buy it.

“The vast majority of consumer-facing assistive gen AI experiences are prematurely launching because everyone has FOMO,” says Emily Pfeiffer, a principal analyst at Forrester who covers commerce technology. “They’re poorly tested.” Industry experts—including those leading agentic commerce efforts at Google and OpenAI—emphasize that change is coming. By the end of this year, they say, AI systems will be able to recommend products and automate checkout with far more ease than I experienced. But once these issues are resolved, there remain deeper, more philosophical hurdles to contend with.

Article truncated for readability. Read the full piece →

The article addresses the significant implications of AI in e-commerce for brand identity and consumer engagement, making it highly relevant and impactful for brand strategy professionals.