Amazon is launching an artificial intelligence tool to help third-party sellers quickly resolve issues with their accounts and retrieve sales and inventory data.

The company said Thursday that it is initially launching the product, called Amelia, in beta for select U.S. sellers before rolling it out to a wider audience later this year. Amazon describes it as an “all-in-one selling expert powered by generative AI” and makes it accessible through Seller Central, the internal dashboard for third-party sellers.

Amelia is the latest generative AI tool that Amazon launched last year to capitalize on the hype surrounding OpenAI’s ChatGPT. The company launched an AI-powered shopping assistant called Rufus, a chatbot for enterprises called Q, and Bedrock, a generative AI service for cloud customers.

As CNBC previously reported, Amazon also plans to upgrade its voice assistant Alexa with generative AI features. The company has also invested billions of dollars in OpenAI competitor Anthropic, its largest venture deal to date.

CEO Andy Jassy told investors earlier this year that the “generative AI opportunity” was almost unprecedented and that increased investment was needed to capitalize on it.

“I don’t know if any of us have seen such an opportunity in technology for a long time, certainly since the cloud, maybe since the internet,” Jassy said on the company’s first-quarter conference call in April.

Google And Microsoft have launched competing products to ensure their relevance in a market whose sales are forecast to exceed the US$1 trillion mark within a decade.

AI is also becoming more prevalent on Amazon’s e-commerce platform. The company now displays AI-generated summaries of product reviews and has introduced AI features for third-party sellers that can help them write listings and generate photos for ads.

Amazon also announced Thursday that it is launching tools that will allow sellers to create AI-generated video ads and use AI to build product listings in bulk based on their entire catalog. The company said it is starting to use generative AI to display personalized product recommendations and listings based on a user’s shopping history. For example, Amazon would display the term “gluten-free” in the description of a cereal box if a shopper typically searches for products with that phrase.

Amazon made the announcements at its annual conference for sellers, held in Seattle. Third-party sellers are the heart of Amazon’s dominant e-commerce business. Since about 2017, they have accounted for at least half of all goods sold on the site. In the second quarter of this year, that number rose to 61%.

Dharmesh Mehta, Amazon’s vice president of worldwide partner services, said in an interview with CNBC that more and more merchants are using the company’s AI services. More than 400,000 of Amazon’s millions of third-party sellers have used the company’s AI listing tool, up from 200,000 in June, he said.

With Amelia, Amazon is using generative AI to help with a key problem for third-party sellers – troubleshooting accounts. The company has sprawling teams to help sellers resolve account lockouts, deal with inventory issues, and build their business on the site. Merchants have long complained that it’s difficult to get a quick resolution or reach a human when unforeseen issues arise with their accounts.

The company said Amelia can help investigate an account issue and will be able to “resolve the issue on the seller’s behalf” in the future. Mehta described how, instead of filling out a missing inventory form, a seller can ask Amelia to file a claim for them or the tool can resolve the issue automatically.

“There will be places where Amelia can maybe do this quicker instead of chatting with seller support or talking to someone on the phone,” Mehta said. “I don’t have to email someone and wait for a response.”

Amazon said Amelia uses Bedrock, a software tool that gives users access to large language models from Amazon and other companies such as Anthropic and Stability AI. Mehta said Amelia is trained using public data from around the internet, along with information from Amazon seller resources, FAQs and other publicly available websites.

Mehta said the model is not trained on seller-specific data, which is kept strictly confidential.

Amazon said the tool uses Retrieval-Augmented Generation (RAG), a popular AI industry framework that combines generative AI with long-established information gathering methods. It allows certain seller-specific information to be retrieved from Amazon’s internal systems without storing it or including it in the model training data.