ChatGPT, Generative AI and How They’re Impacting Retail

ChatGPT screen feature

ChatGPT from OpenAI is causing serious disruption, including in retail (photo credit: Om siva Prakash on Unsplash)

ChatGPT, the generative AI tool based on Microsoft-backed OpenAI technology, is creating media buzz on an entirely different level, with its much-touted ability to create human-seeming language, content and interactions for just about any business or private context.

Like many little-known technologies, it’s also provoking a lot of fear and suspicion. Concerns range from cybersecurity, intellectual property rights, liability exposure and ethics – including the jaw-dropping irony of using it to cheat in an AI ethics class – with opposition from a wide range of groups. For instance, many educators are now suddenly forced to find ways to counter GPT-driven plagiarism, which even the CEO of OpenAI admits can’t be completely thwarted.

Not only, that, ChatGPT, in the guise of a Bing chatbot, has been known to throw shade and react nastily when confronted with the fact that its information was wrong, according to The Verge. It even allegedly claimed to have spied on Microsoft’s own developers through laptop webcams, the publication reported.

On the retail front, the capability is enabling companies like Shein to completely reinvent the term “fast fashion.” The Chinese giant churns out new apparel items in a couple of weeks to instantly cash in on a hot TikTok trend, exactly matching production of each new SKU to demand to avoid excess inventory, reduce costs and maximize profit.

Not a New Thing, But Now a Force of Nature

The concept of generative AI, i.e., using algorithms to generate new content, is not a new one, having been used in retail and ecommerce contexts for years. And AI in retail itself goes back even farther, to Amazon’s pioneering work on algorithm-driven recommendation engines in the early 2000s.

But ChatGPT has burst on the scene late last year like a hurricane, and there’s now an arms race between it and Alphabet/Google’s Bard, a competing tool. Meta, meanwhile, had an embarrassing three-day false start in November with its Galactica large language model for scientists.

“I don’t think a lot of retailers are panicking because our world is suddenly different (with ChatGPT),” said Jason Golberg, chief commerce strategy officer at Publicis. “They all knew the world was different before this became a trend six weeks ago.”

Other examples of generative AI capabilities include competitors Stable Diffusion and Dall-E for creating images where none existed, and, which has licensed ChatGPT3 to auto-generate any and all kinds of copy – including journalism (editor’s note: I researched and wrote this whole thing).

Various Applications for Retail

“Imagine you’re Walmart, trying to sell pants from this vendor, a shirt from that,” Goldberg said. “They can use generative AI to create an image of a 40-year-old man, weighing 200 lbs. wearing both. They have a usable product image for selling they didn’t have before.”

In another application, AI is used for sizing solutions that allow shoppers to virtually try on clothing for fit before buying, saving money by reducing returns. The use cases are endless, Golberg said, and AI enables companies to do things at scale that weren’t possible before.

Generative AI’s ability to create instant product descriptions using some basic inputs has enabled Goodwill to stand up its ecommerce operation, with countless one-off SKUs making it impossible to do so manually. “Goodwill uses AI to take a picture of that Members Only bomber jacket and writes a whole description, saving them hundreds in labor on the product detail page,” Goldberg said.

Experiments, Issues and Some Caution

Bharath Thota, a partner at Kearney covering advanced analytics and data science, said lots of retailers are asking him about ChatGPT and its capabilities, with many in experimentation mode. His advice is to take a measured approach, for several reasons. For one thing, Thota said, it takes a lot of computer processing power to handle the amount of data needed to effectively train the algorithms used in generative AI.

Adding another layer of challenge is the growing number of technology providers hawking various generative AI tools, Thota said. They’re knocking on the doors of retailers and making their presence felt at major trade events like Shoptalk and NRF’s Big Show, where AI-based solutions abounded last month.

“A lot of companies still struggle to build the foundational data layer, and the way ChatGPT works, it learns from what you feed it – it’s garbage in, garbage out,” he said. “There’ll be a lot more emphasis on the quality of data and processing power. Companies may not be ready to invest in those types of high-end machines. So, there are a lot of things that need to happen from a foundational layer perspective in order to see the real fruits of GPT and all these generative AI models to come to life.”

Thota also said in the case of generative AI’s ability to quickly spin out faster and faster fashion a la Shein, most apparel brands and retailers are not built that way. The product design process involves review and approval from several internal teams, starting with manufacturing and marketing – and it can’t be changed overnight.

“Imagine you’re getting instant feedback on what customers actually like about your product and what they don’t like,” he said. “But you have an internal process set up for all these teams to come together and work with each other to align on a new feature or a design change. That also needs to be changed and addressed for you to embrace the output coming from the AI systems.”

Beyond that, Thota said, corporate governance teams aren’t equipped to deal with the real-time flow of generative content and the possible liability issues it could raise from sensitive or offensive material.

“We’re still trying to figure out things about copyright infringement, IP protection and the legalities of the content being produced,” he said. “Typically, when a human does it, there’s a legal policy standard and a governance team that looks at everything. But now you need a new kind of governance model that looks at what the system is generating vs. humans, and nobody has cracked the code on how to set that up.”

One Advantage, And Different Approaches

While Shein’s goods might be bad for the environment and “made in awful factories in China,” Goldberg said, the suddenly ascendent retailer has an advantage over established apparel companies: greater inventory precision not dependent on long-range forecasts.

“In one way, they’re wildly better,” he said. “A major apparel retailer might sell half of what they produce, because of a bad match of supply with demand, so they liquidate the other half. How much does Shein make and sell? All of it. They look at a demand curve that’s a week old, and they’re done.”

Many retailers are trying to figure out how to move faster with generative AI, while others are pivoting to less-trendy products that are more evergreen, Goldberg said. “Gap Inc. is not chasing the fast fashion trend,” he said. “They’re selling staples, solid colors, and not worrying about what’s everyone is saying on TikTok. Some are banking on enough big slow-moving trends to stay in existence while all this changes around them.”