New AI shopping agents promise better bargains. But who pays the price?

Meta’s recently launched AI agent Muse promises to be many things: it can fire off emails, book holidays and fill out cumbersome online forms.

But it’s attracting particular attention for its potential as a personal shopper that can search for bargains and make purchases without a user’s approval.

Just think of someone hunting for a new pair of running shoes. Within minutes, an AI assistant could scan dozens of retailers and compare different brands to find the cheapest suitable pair.

And Muse is far from alone. Google and OpenAI are also developing AI-powered shopping tools, as tech giants compete to make these agents part of everyday purchasing.

The appeal of these new AI agents to consumers is obvious: less time spent shopping, and potentially better deals. For retailers, they could change how products are marketed and recommended online.

Yet, there’s another important implication getting less attention: what happens to the workers who make those products? And will they be left even worse off?

The workers we can’t see

Major global brands negotiate prices with suppliers, which also must manage their own production costs. Materials might be sourced in one country, components manufactured in another and finished products assembled elsewhere. These layers can make workers – and the conditions they operate in – invisible.

The consequences of buyers’ demands for workers are well documented. Research by the International Labour Organisation involving 31 suppliers across five countries found that buyers’ demands put pressure on wages, increased overtime and sometimes led factories to cut corners on workplace safety.

Indeed, UK researchers have demonstrated how consumer relationships and global supply chains influence the governance of labour standards. More recently, Australian labour policy expert David Peetz has explored how interconnected technologies can concentrate economic power.

Millions of AI shopping agents scouring the internet for the lowest prices could intensify these pressures, particularly if algorithms increasingly favour cheaper products. Some businesses might respond through innovation, but others could look for savings by cutting labour costs.

For workers vulnerable to exploitation, the consequences could be more serious. Worldwide, tens of millions of people already live in modern slavery. Forced labour remains a major problem in global supply chains, particularly in industries such as manufacturing and agriculture.

Such exploitation isn’t simply the result of unethical employers. Pressure to cut costs, weak regulation and workers’ limited bargaining power can all to poor working conditions.

So could consumers simply tell their AI shopping agents to buy ethically instead?

The difficulty is knowing which products meet that standard. Expensive running shoes may cost more because of branding or profit margins, rather than better wages or working conditions. And companies’ ethical sourcing claims may reveal little about what happens in subcontracted factories.

An AI assistant can readily compare prices, delivery times and customer reviews. But it has a much harder task establishing whether workers are paid fairly, can join unions or have safe workplaces – information even human consumers struggle to obtain.

There is also a larger economic contradiction. AI shopping promises to increase our purchasing power by helping us find better products, avoid inflated prices and save money.

But consumers are also workers. If cheaper goods come at the expense of wages and job security, some of the people enjoying those bargains may ultimately have less money to spend. What they gain at the checkout could be offset by what they lose at work.

Could AI make ethical shopping easier?

It’s also possible AI shopping agents could help consumers make more informed choices – but only if they have access to reliable information about how products are made.

Some of that information is already being collected. Open Supply Hub, for example, maintains a public database mapping more than 2.5 million production locations worldwide, helping identify factories and the companies connected to them.

Imagine an AI shopping assistant using such information, alongside independently verified labour standards, to find running shoes made under fair working conditions rather than simply recommending the cheapest pair.

But knowing where a product was made is not the same as knowing how workers were treated. That requires independent scrutiny, information from workers themselves and effective protections for their rights.

The OECD already has guidelines for responsible sourcing in the clothing and footwear industries, while the International Labour Organisation’s Better Work programme has demonstrated that efforts to improve factory conditions can produce results.

AI could make this information easier for consumers to find and compare. But it can’t replace the regulation, collective bargaining and accountability needed to improve working conditions in the first place.

Tracing a product’s whakapapa – its origins and connections – means recognising the people and employment relationships behind an everyday purchase.

Whether AI makes those connections more visible, or obscures them further, will depend partly on the information available and what shopping agents are designed to prioritise.