Some retailers are using AI to churn out hundreds of product descriptions that previously took weeks to create.
oh i want to update For your work from home wardrobe. You go online and start browsing basic information on a daily basis. Choose from striped tees, scoop neck tees, pocket tees, graphic tees, v-neck tees, button back tees and more. And it’s just a t-shirt.
In the past, besieged copywriters have had to muster their energy to spin out a passionate description of each piece, while being careful to accurately describe materials and fit. It may take several weeks. Now, some retailers are testing whether they can use generative artificial intelligence instead, using the same underlying technology as ChatGPT.
Last year, Stitch Fix began using AI to create product descriptions for its online catalog of thousands of products. “In 30 minutes he can write 10,000 product descriptions,” said Sachin Dhawan, chief technology officer at Stitch Fix. It usually takes him less than a minute to review a copy for an employee. Copy he gets approved 9 times out of 10. “They are very good,” he added.
Retailers are looking to this technology as a way to reduce the time and expense of running e-commerce operations, especially as profits are being squeezed by high costs and rising discount levels. According to ZipRecruiter, the average copywriter’s hourly wage is $29, or about $60,000 per year. AI can do the job in a fraction of the time and at a low cost. In a Coresight Research survey, more than a third of his retailers say AI helps him create product descriptions that represent the voice of the brand.
At lingerie brand Adore Me, which was acquired by Victoria’s Secret, a copywriter complained that writing product descriptions was one of her least favorite jobs, taking 20 to 30 hours a month. “It’s not the most fun when he has to try to be original, fun and creative 100 times in a row,” said Ranjan Roy, Adore Me’s vice president of strategy. increase.
You can now use AI to revitalize your processes. “It’s better than staring at a blank screen,” says Brian Hennessey, CEO of Talkoot, which uses AI to help online retailers speed up copy. Companies may end up rewriting 80% of the most prominent and important product descriptions, he said, but AI could at least help get that started. Volume driver relies heavily on AI and may only change his 20% of copies.
“A sock is a sock. A white t-shirt is a white t-shirt,” said Hennessy, formerly global lighting director for Adidas. “I know what they are. AI knows what it is.”
Hennessey’s customers, including Adidas, Under Armor, Reebok, and Burton, have saved 40% in content production, reduced errors by 3x, and increased conversion rates by 10%.
Snowboard company Ulton uses AI to personalize the shopping experience. The company knows that its website attracts more hardcore snowboarders, while its Amazon pages tend to attract parents and beginners. So the company uses his AI to create personalized product descriptions, publishes more insider language on its website, and more descriptive and educational language on Amazon.
AI is also great at optimizing language for searches. Total non-branded searches for Adore Me increased by 40%. This means that when someone unfamiliar with the brand does a general search online for bras, underwear, swimwear, etc., it will appear higher in search results.
Algorithmic salesmanship takes time. When Adore Me first started working on this technology, its descriptions were generic and off-brand, using terms that humans probably wouldn’t use, like “while”. The company then began training the algorithm on its own data, starting with him inputting all of his 23,000 product descriptions the company has created since its inception 10 years ago.
“Bigger is better,” says Roy of Adore Me, but it contained all sorts of contradictions by different writers in different eras and different styles.
Ultimately, the brand supplied copies of the last two years that had a more consistent brand voice, and tweaked from there. For compliance purposes, we instructed the algorithm not to include certain words such as “green” and “environmentally friendly”. We made sure each description was optimized for search, including always including references to matching bras and panties.
talk the talk
Adore Me spent months training an AI to write product descriptions that nailed the brand’s unique style and tone.
Take 1 — includes terms that are generic, off-brand, and never used (while, wiring)
Our Rylie Black Push Up Balconette Underwired Bra is the perfect combination of comfort and great shape. Soft molded cups are lined with cotton for breathability, and pretty floral underwires provide great support. A hook and eye back closure for added security and fully adjustable straps allow for adjustment both around the shoulders and back.
Take 2 — Trained on unique product descriptions, creative and fun, but not optimized for search
A little black number that takes you from your desk to the dance floor. The Push-Up Balconette Riley features a lace-trimmed neckline and contrast bow for a sleek, modern look. Complete the style with matching panties.
Take 3 — Trained on a narrower range of product descriptions and optimized for search
This tuxedo-inspired black push-up brings scalloped panels tailored to accentuate the figure, while a balconette design and subtle push-ups ensure support. Perfect for a bold, modern look, pair it with a matching thong and hipster panties to feel as comfortable as it looks. (There are sizes from 30A to 38DD.)
“That’s crazy,” said May Habib, CEO of AI company Writer, which worked with Adore Me on the product description. “It can almost completely eliminate human effort in the writing process.”
Adore Me has not laid off its copywriters, but said their job descriptions have changed to focus on more sophisticated marketing strategies and creative campaign work.
Stitch Fix also said the copywriting team remained intact. Despite spending years training algorithms to take over more functions, humans are always “informed” and able to make decisions. For example, the stylist reviews computer-generated clothing recommendations before sending them to shoppers who sign up for a subscription box. But its à la carte product, called Freestyle, independently generates recommendations in near-real time. All of which will help reduce your workload. It’s hard to say how much the algorithm is to blame, but the company’s annual report shows that the number of stylists employed by Stitch Fix has dropped by a third since 2019, while the number of data scientists and engineers has fallen. Consider that it’s increased by a third. .
More companies are moving in this direction. Large retailers and food and beverage companies are most interested in AI-generated copy, Hennessey said. He said that following the recent buzz around ChatGPT, some companies are ready to fully embrace the technology and face high expectations.
“It’s like coming to a farmer who has been plowing behind his mule for 40 years. And I said to the farmer, ‘Do I still have to drive?'” Hennessy said. “We are trying to counteract the idea that AI can do anything.”