Article

Why buyers refuse to let AI agents spend their money

Published November 19, 2026 6 min read By the Lotvs team 1 sources, read November 19, 2026

The retail industry is building for a future that buyers are actively rejecting. Everyone is rushing to build AI agents that can sell, negotiate, and execute transactions. But a massive gap exists between what the industry is building and what people actually want to use.

On November 4, 2026, Talkdesk released a survey of 1,048 US shoppers and 312 retail leaders. The numbers reveal a stark disconnect. Retailers are betting heavily on agentic commerce. A full 89 percent of them are preparing for AI agents to shop on a customer's behalf. They envision a world where machines talk to machines to execute transactions seamlessly.

The buyers are refusing to participate in that vision. Only 15 percent of shoppers are willing to let an AI agent make purchases using their payment details. The industry is building autonomous spenders. The public is demanding strict supervisors.

The boundary between research and action

People are not rejecting AI. They are rejecting autonomous spending. The Talkdesk data shows that 74 percent of shoppers plan to use AI assistants for shopping this season. They want the machine to find the deals. They want it to compare specifications across dozens of websites. They are even open to new suggestions. Nearly half of the respondents, 49 percent, are willing to buy from unfamiliar brands based on an AI recommendation.

The machine is an excellent researcher. It processes vast amounts of data faster than a human ever could. But research is where the delegation ends. When the task shifts from gathering information to executing a trade, the trust vanishes.

Even if they are allowed to review the transaction beforehand, only 15 percent of buyers will let the agent use their payment details. If you remove that review step, the number falls to 10 percent. Nine out of ten people absolutely refuse to let a machine spend their money without a leash. They want the agent to do the heavy lifting of discovery. They demand to be the ones who click approve.

The cost of autonomous mistakes

Why is the public so hesitant? The answer lies in the performance of the AI tools already in the market. Retailers are rushing to deploy AI in customer service, with 93 percent doing so this season. But these deployments are flawed.

Nearly one in three retailers using AI, 31 percent, report their tool took an incorrect action on a customer account in the past twelve months. This includes serious financial errors like issuing an improper refund. The overall error rate is even higher. More than half of the retailers, 58 percent, reported AI errors.

These are not victimless mistakes. Among the retailers that reported errors, 29 percent lost a customer as a direct result. Other problems include customer complaints about AI interactions at 33 percent, inaccurate product recommendations at 25 percent, and biased or inappropriate outputs at 22 percent.

When an agent hallucinates a product feature, it is an annoyance. When an agent hallucinates a price or executes a flawed transaction, it is a material loss. Buyers understand this risk intuitively. They know the technology is prone to errors. They will not hand over their digital dollars to a system that might accidentally buy the wrong item, authorize an incorrect charge, or fall for a deceptive listing.

The transparency deficit

Trust requires clear rules and absolute transparency. The retail industry is currently failing on both fronts. More than three quarters of shoppers, 77 percent, want to know when AI is part of their shopping experience. Yet 37 percent of retailers do not fully disclose it.

This lack of transparency breeds deep suspicion. Shoppers have specific, well founded fears about how AI will be used to manipulate them. Among the 74 percent of shoppers demanding stronger AI rules, 55 percent worry about unauthorized personal data collection. They fear the agent is harvesting their habits rather than helping them shop.

Another 41 percent worry about paid product placement. They suspect the agent is recommending an item not because it is the best fit, but because the manufacturer paid for the placement. Most concerning for the future of agentic commerce, 41 percent worry about being charged a different price than another shopper. They fear the AI will use their data to calculate the maximum amount they are willing to pay and adjust the price accordingly.

If buyers suspect the agent is working for the retailer's margin rather than for their own benefit, they will never grant it spending authority.

What changes for builders

This data provides a clear roadmap for anyone building in the agent economy. You cannot build autonomous spending loops. You must build approval loops. The user must remain the final authority on every transaction. The agent proposes. The human disposes.

This requires a fundamental shift in how we design agentic systems. We must separate the intent to buy from the execution of the payment. When an agent finds the right item, it cannot simply pass a saved card number to a checkout API. It must halt the process and request permission.

Agent payment approval flow. An AI agent finds a product, requests payment, and waits for the human user to review and approve the transaction.

The protocol level is where this separation happens. When an agent attempts to access a paid resource or initiate a checkout, the merchant server should reject the initial request and demand payment.

HTTP
HTTP/1.1 402 Payment Required
Content-Type: application/json
Agent-Payment-Request: true
Amount: 45.00
Currency: USD

{
  "error": "Payment required to complete this order",
  "checkout_url": "https://api.merchant.com/checkout/12345",
  "expires_in": 600
}

The agent catches this requirement. It cannot fulfill the requirement on its own because it holds no funds and possesses no spending authority. Instead, it parses the request and surfaces it to the human user. The interface must clearly display the merchant name, the exact amount, and the item details.

The human reviews this information. Only then does the human approve the transaction and authorize the release of digital dollars. The agent then takes that signed approval back to the merchant to complete the loop. This is the only path forward. As Talkdesk notes, "trust is the ultimate currency" in this new environment. Earning that trust requires hard limits.

The future of seasonal work

It is also worth noting how this shift affects the human workforce behind the scenes. The rise of AI agents is not leading to massive job cuts in retail. Instead, it is changing the nature of the work.

Nearly half of retailers, 47 percent, plan to hire approximately the same number of seasonal workers as last year. Only 19 percent plan to reduce seasonal hiring due to AI. Retailers are building a hybrid workforce. They intend to use AI agents for specific tasks, with 51 percent planning to use them for returns, refunds, or exchanges.

The human workers will handle the complex cases that require empathy and judgment. To prepare for this, 89 percent of retailers plan to upskill or reskill their workers. The machines will handle the volume. The humans will handle the exceptions.

The path forward

Agentic commerce is the next major shift in how we interact with the internet. Retailers are spending heavily to prepare for it. But the transition will not happen by default. Buyers will not surrender control of their money to autonomous systems. They demand transparency, they fear mistakes, and they insist on having the final say. The builders who succeed will be the ones who respect these boundaries. They will build agents that are brilliant researchers and obedient executors. They will build systems where the human always clicks approve.

Share this article

Next step

Try it with Lotvs

Connect Claude, ChatGPT or any MCP client to an account of its own, with limits you set.

Sources

Every fact above links to one of these. Each was read on November 19, 2026; the date beside it is the page's own.

  1. Talkdesk, talkdesk.com, November 4, 2026

More from the blog

Guides

  • AI agents and your accountConnect Claude, ChatGPT or any MCP client to your Lotvs account. It can read your balance and activity, tell you what it finds and ask you to pay someone or to convert. Nothing moves until you approve it.
  • Payments your AI preparesYour AI agent can ask you to pay someone in digital dollars (USDC) or digital euros (EURC). Each payment waits in Lotvs until you approve it and sign it from your own account. No Lotvs fee to send.

Important

Articles describe products and protocols as their sources did on the dates shown, and those change; check the sources before you rely on them. An AI agent connected to Lotvs has no money of its own. It can read your account, tell you what it finds and ask you to pay someone or to make a conversion. Nothing moves until you approve it in Lotvs and sign it from your own account.

Lotvs is a financial technology company, not a bank. Balances are held as digital dollars, not bank deposits, and are not covered by deposit insurance. Names of other companies and their products belong to them, and Lotvs is not affiliated with them. See the Terms and Privacy Policy.