The Race for AI Shopping Agents

This holiday season, millions of Americans will fire up their laptops to shop for gifts. However, tech companies are racing to automate this process using AI agents.

Recently, Perplexity introduced an AI shopping assistant for its subscribers in the U.S. This agent is designed to browse retail websites, locate the desired products, and even complete the checkout process for users.

While Perplexity might be the first major AI company to launch such a service, others have been active in this area for some time. Expect to see more AI shopping agents in 2025. Tech giants like Google and OpenAI are reportedly working on their own versions of AI agents capable of making purchases, including booking flights and hotels.

Big players like Amazon could also enhance their AI chatbot, Rufus, to facilitate checkout, as it would be a natural evolution given the volume of product searches conducted on its platform.

Companies are utilizing a combination of innovative and traditional approaches to overcome obstacles set by retailers who wish to block undesired bot activity. For instance, Rabbit launched its LAM Playground, allowing AI agents to navigate websites through virtual machines in data centers. Anthropic also provides a similar service, but hosted directly on your personal computer.

Additionally, Perplexity has teamed up with Stripe to utilize some older payment methods adapted for AI agents. Stripe is providing single-use debit cards that Perplexity’s AI shopping agent can use for online purchases. This limits the agent's spending; if it makes a mistake, like buying the wrong item, only a small amount is lost instead of accessing your entire bank account.

Reports indicate that Google’s AI agent will need credit card access, which could raise privacy concerns for consumers. However, familiar entities like Google, Amazon, and Apple already have access to users' billing information and often fill out shopping forms automatically. This could give them an edge in the upcoming competitive landscape.

These new tools could significantly alter online shopping, potentially upsetting retailers and advertisers who currently profit from the conventional system.

Like AI chatbots that efficiently find hard-to-locate information, AI shopping agents may help consumers discover products or deals they might otherwise miss. In theory, these tools could save users a considerable amount of time when searching for budget-friendly travel options or suitable gifts.

Yet, it’s clear that we have a long way to go before AI agents can fulfill all the holiday shopping needs. Despite their early designs, these agents are in an experimental phase.

According to initial evaluations, Perplexity’s shopping agent can take several hours to process purchases, and it occasionally fails to secure items altogether. As it stands, using this agent can feel more cumbersome than traditional shopping on platforms like Amazon.

Perplexity has also included human monitors in the process to ensure its AI agent operates correctly. While human oversight is becoming more common in AI systems, it's noteworthy that these human checkers may have access to private information like purchase items and billing addresses, raising potential privacy concerns.

In one test of Perplexity's shopping agent, the task was to buy toothpaste. Upon entering the request, the agent provided options from Walmart and Amazon, among others. When selecting a tube of Crest from Walmart, it looked like I could seamlessly checkout from the Perplexity platform. However, my bank statement reflected that I had paid Perplexity, not Walmart. Eventually, three hours later, Perplexity informed me that the toothpaste was sold out. The next attempt took eight hours before I received a purchase confirmation.

This raises several questions about the efficiency and reliability of these AI shopping agents. Perplexity’s approach involves scraping retail websites, which means there is often a delay in real-time updates. This discrepancy can lead to confusion between what the agent indicates is available and the actual inventory.

There may also be an unauthorized element to how Perplexity utilizes the retailers’ listings, which could complicate both purchasing and returning processes.

When a user checks out through Perplexity, they are essentially paying Perplexity the exact cost of the item, providing instructions for the AI to make a purchase, and submitting their shipping details. This task might take some time to complete, with the AI agent attempting to fulfill the order after a delay.

Stripe’s product lead noted that using an AI agent is similar to handing a small amount of cash to a personal assistant with specific spending rules. However, the AI requires constant oversight, necessitating human input for monitoring and ensuring accurate transactions.

As AI shopping agents gain popularity, it could mean fewer consumers visiting online shops, which could impact retailers' abilities to suggest additional products and perform upsells. Advertisers might also lose insights into customer behavior, leading to fewer targeting opportunities.

Because of these potential shifts, retailers and marketers may resist the onset of AI shopping agents. This is why companies like Rabbit and Anthropic are developing AI systems that interact with retail websites just as a human would, respecting the typical user interface. This strategy helps avoid needing prior permission from companies to access online services that could restrict their bot usage.

Rabbit's CEO recently highlighted that AI agents are becoming increasingly skilled at solving CAPTCHA tests, which have typically hindered bots from conducting online purchases. Therefore, online retailers may need to enhance their verification measures to maintain security.

Ultimately, it is feasible that AI agents could provide a superior online shopping experience in the future. While the current iteration of Perplexity’s shopping agent is far from perfect, it signifies a step towards that potential.

In the coming year, improvements to AI shopping agents from companies like Perplexity, OpenAI, and Google might lead to more advanced features, changing the landscape of online retail significantly and highlighting the challenges faced by AI agent developers.

shopping, AI, technology