From documents leaked to Insider, Amazon is launching a secret project codenamed “Project Nile” to revamp its search experience on its website and app with AI-powered chat features. This project is being spearheaded by Amazon VP Joseph Sirosh, an AI executive who joined Amazon in 2022 from Microsoft.
Project Nile’s goal is to create a conversational shopping agent, mimicking an in-store salesperson, to provide instant product comparisons, expert advice, and personalized recommendations.
The project is expected to launch in the US in January, starting with the mobile app. Sirosh aims to improve mobile sales conversion, as mobile search constitutes a significant volume but has a lower conversion rate than the desktop version.
Amazon isn’t necessarily going it alone as they develop Project Nile. They are open to partnering with other AI companies to use existing tools such as ChatGPT and YouChat to help provide answers to sensitive questions or provide data support.
![](https://vasogroup.com/wp-content/uploads/2023/10/AdobeStock_370280312-1024x567.jpeg)
Potential Issues
Amazon’s ambition to dramatically transform its search business comes amid intensifying competition and potential restrictions related to data security and regulatory compliance, highlighting the need to balance innovation with user protection.
Ensuring data privacy and security is key. Project Nile will hinge on the collection and processing of significant amounts of user data, including search history, personal preferences, and purchase behavior. Any partnership with a third-party provider will only complicate this.
Accuracy and trustworthiness is a concern, as AI systems are known to generate erroneous information. Amazon plans to employ human AI trainers who will review AI-generated answers, ensuring accuracy and addressing any false info that slips through.
In the backdrop of Project Nile is a pending FTC lawsuit alleging anticompetitive practices by Amazon. The outcome of that legal battle could also impact Project Nile’s operations.
Our Assessment
From a content perspective, Project Nile reaffirms the necessity of providing accurate, abundant information to the shopper. That, along with providing all available attributes on set up will help fuel the dataset.
Shoppers clearly prefer visual information over text, and Amazon will continue to prioritize product listings with high-resolution images and video within its current algorithm. Amazon may further enhance this trend by utilizing AI to index infographics and videos in its search results.
Project Nile will also heavily rely on user reviews and customer Q&As. It will be essential for vendors to have a solid customer relationship management (CRM) system in place, and to ensure they are using any allowed means possible to boost their reviews (e.g. Vine, Early Reviewer Program, product inserts).
From the advertising perspective, this could pave the way for significantly more precise targeting. To deliver tailored recommendations, Project Nile’s chatbot salesperson will need an intricate comprehension of product listings that doesn’t currently exist. A comprehensive AI dataset could potentially translate into a comprehensive dataset for advertising campaigns.
Currently, product attribute targeting (PAT) enables us to target ads based on comparative price, comparative reviews, etc. Using AI to expand comparative targeting (material, color, fit, etc.) is a logical next step to increase ad relevance for customers.
Changing Behavior
One issue we don’t see addressed in the original reporting: Shopper behavior.
In order for Project Nile to work shoppers must change their search behavior on Amazon. Currently it is difficult to envision Amazon customers typing, “What kind of coffee maker should I get?” rather than specifying their actual preferences, such as “Keurig coffee maker,” “5 cup coffee maker,” or “cold brew coffee maker.”
Amazon shoppers typically exhibit focused shopping intent, often arriving with a clear idea of what they intend to purchase. Broad, top-of-funnel questions are typically associated with Google. Google has made many unsuccessful attempts to emulate Amazon’s e-commerce success. Will Amazon succeed at Google’s top-of-funnel game?
For real change, Amazon is looking to artificial intelligence.