How Autonomous AI Shopping Agents Will Transform Retail

25 Feb 2025

Retail's AI revolution is entering its third wave. After predictive AI and generative AI, autonomous agents capable of completing shopping tasks without human intervention are emerging as the next frontier. Salesforce's latest industry research reveals that 32% of consumer goods companies have already fully implemented generative AI, with digital commerce as a primary focus area. As the technology evolves from answering questions to taking action, brands and retailers face urgent decisions about how to adapt their digital presence, product content, and media strategies.

The transition from generative AI to agentic AI represents a fundamental shift in capabilities. While chatbots and assistants like Amazon’s Rufus can answer questions about products, autonomous agents can complete entire shopping journeys—from discovery to purchase—with minimal human intervention.

AI’s Evolution In The Consumer Goods Industry
According to the joint Salesforce and Accenture "Industry Insights Report: AI Edition," we’re witnessing a clear progression in AI capability.
Michelle Grant, Director of Strategy and Insights for Retail and Consumer Goods at Salesforce, offers a helpful distinction between traditional automation that marketers are likely familiar with and newer AI approaches:

Traditional Automation follows predefined steps, but it’s not artificial intelligence nor is it agentic. For example, if someone opens an email, they’re automatically added to 'Group A' and, if not, a follow-up email is sent. It doesn't analyze data, make decisions, or learn over time–it's rules-based automation.

Predictive AI (Wave 1) uses historical and statistical data models to predict the future. For example, predictive AI uses machine learning algorithms to analyze a shopper’s historical engagement data and predict the best time to send an email.

Generative AI (Wave 2) is used for creating new content using LLMs and data. Examples include summaries, text generators, and image generators based on prompts. While it can produce content, it doesn't independently make decisions or take action.

Agentic AI (Wave 3) uses machine learning and natural language processing to autonomously get work done, without requiring human input.

Grant explains that the key difference here is that agentic AI can take action based on its inputs to do things like send a generated email, develop campaign strategies from its insights, or add products to carts based on shopper preferences.

For retailers and brands, this progression isn't merely academic—it's reshaping how consumers discover and purchase products. Consumer goods companies are already identifying their most valuable AI agent use cases, with "helping shoppers find products on websites or other digital platforms" ranking third in priority.

From Answering Questions To Taking Action

The distinction between generative and agentic AI becomes clearer when examining real-world implementations. Saks, for example, launched Agentforce (Salesforce's agentic AI platform) in September 2024 to enhance its customer experience.

Within its Saks Chatbot, Agentforce analyzes customer interactions and determines the next best action based on the context, while automating and streamlining tasks and inquiries.

A video demo of Saks’ Agentforce integration shows an SMS interaction between a customer and Saks’ AI agent, whereby the customer shares some photos of an outfit inspiration and Saks’ agent returns with similar items. It knows her usual size, just as a personal sylist would, and helps to coordinate an order and later size swap.

SharkNinja, a global product design and technology company behind the Shark and Ninja brands, is also implementing Agentforce to enable SharkNinja to easily build and deploy AI agents that can autonomously take action across any business function, the company says. With Agentforce, SharkNinja will have an always-on, digital workforce available 24/7 to guide customers through the buying process, answer product questions, troubleshoot issues, and manage returns.

Transforming Retail Media With Agentic AI
For retail media networks, the rise of AI shopping agents creates both challenges and opportunities. Currently, retail media spending is heavily skewed toward lower-funnel conversions—in a recent analysis I covered on retail media budgets, over 71% of spending occurs in sponsored products or similar bottom-funnel placements.

But what happens when AI agents, not humans, are making or influencing purchasing decisions? The traditional emphasis on eye-catching creative and emotional triggers may give way to more structured, attribute-based approaches that persuade algorithms, not people.

Take Walmart Connect or Amazon Advertising, for instance. Brands currently bid on keywords and placements to capture consumer attention. In an agentic AI world, they may instead need to optimize for the parameters and ranking factors that AI shopping agents prioritize—potentially shifting spending from traditional sponsored product ads to digital shelf optimization and structured data initiatives.

According to the Salesforce/Accenture report, the second-most beneficial AI agent use case involves optimizing marketing and advertising campaigns. Several technology companies are already addressing this need by developing AI platforms that can enhance campaign management. Xnurta, for example, is an AI-powered ad management platform that enhances campaign management on Amazon and Walmart by predicting buying patterns and optimizing in real-time (disclaimer: Xnurta is a client of mine).

As agentic AI evolves, these platforms will likely expand from optimization tools to autonomous marketing agents capable of managing entire campaigns with minimal human oversight.