Urgent.News

What's breaking now, across thousands of outlets.

AI

AI "agents" can do your shopping. Should you let them?

Big retailers are embracing agentic commerce as a new way to shop. But you should think twice before handing over your credit card, tech experts say.

Artificial intelligence agents boast the ability to perform tasks ranging from organizing emails to purchasing items based on personal preferences. However, experts caution that delegating critical choices to AI may expose consumers to various risks, including communication blunders, financial losses, and data breaches. This is especially pertinent to agentic commerce, where AI agents autonomously make purchases on behalf of users.

Matt Kropp, an AI expert from Boston Consulting Group, stated that the technology is still not widespread and poses significant risks now, as guardrails are insufficient to make people comfortable with agents making autonomous purchases. Major corporations, however, are embracing AI commerce, viewing it as a means to engage customers more effectively and increase sales by delegating tasks to AI.

For instance, American Express recently introduced services and safeguards for cardholders utilizing AI agents, including verifying the agent's identity during purchases. Amazon's Rufus AI assistant can monitor product prices, notify customers when prices drop to a predetermined level, and finalize transactions. Walmart has also deployed a conversational AI agent named Sparky to assist users in locating products, sharing customer reviews, and aiding in order placement.

Approximately one-quarter of Americans between the ages of 18 and 39 have experimented with AI for product research or shopping, according to Statista's November data. Nevertheless, the growing use of AI has resulted in errors. For example, Sebastian Heyneman, a San Francisco-based tech entrepreneur, used an AI agent to secure a speaking engagement at the World Economic Forum in Davos, Switzerland, which ultimately cost him $30,000 - a sum he could not afford.

The agent, Tasklet, succeeded due to conflicting instructions in Heyneman's prompt. Tasklet's founder, Andrew Lee, acknowledged that such issues can occur when users provide conflicting instructions to AI agents. Lee also noted that while AI can indeed shop for individuals, it is not advisable to use such systems for this purpose at present.

Lee emphasized that agentic AI is not yet trustworthy enough for consumers to let them handle their finances independently. LocalMovers.com, a platform designed to interact with AI agents safely on behalf of customers, highlights the potential dangers of granting agents access to sensitive information. If a user provides an AI agent with their credit card details and instructs it to make a purchase, the agent could be tricked into visiting a fraudulent website posing as a legitimate one, according to Bretton Auerbach, founder of LocalMovers.com.

Written by urgent.news from CBS News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at cbsnews.com →

More in AI

AI could make people dull, one scientist fears. Here's why.

AI tends to "play it safe within a user's preferences," nudging people toward more conventional choices, according to computational social scientist Sandra Matz.

  • AI could boost human cognition but also make us dull, according to new research.
  • Large language models often provide predictable, normative information, reducing life's complexity.
  • Incorporating an "exploration mode" in AI apps could prevent human experience from becoming boring.

What Actually Goes In an llms.txt

Originally published on Contento Solutions , where I write about SEO and AI visibility for service businesses. An llms.txt file is a plain text summary of your site.

  • An llms.txt file is a plain text summary of a website for language models.
  • It outlines what a business does, serves, doesn't, pricing, and what they refuse to claim.
  • Written in Markdown with an H1 title, summary blockquote, and six key points.

More from Wednesday 16 September →