AI shopping bots: faster and cheaper deals?

Is an AI bot a faster and cheaper way to shop?

Personal preferences are more important than the bot you use.

Today, we all shop online far more often than before the pandemic. It’s no surprise therefore that advertisers are competing fiercely for our digital attention. However, this could change very quickly. In the near future, will it be our personal AI assistant wading through adverts and clickbait in search of the perfect bargain? This would certainly save us time and money. But would we all really be better off?

Negotiations on autopilot: who would actually win?

Researchers from the universities of Chicago and Michigan explored what AI-assisted shopping might look like.
 
Their conclusion is surprising. AI negotiations can lead to extremely different outcomes, and this has little to do with the bot itself and everything to do with the person controlling it.

The researchers asked participants with varying degrees of 'machine fluency' (the ability to communicate effectively with AI) to use negotiation bots for straightforward transactions, such as purchasing a second-hand item or booking a hotel room.

Those who could formulate precise and strategic prompts – for instance, by instructing the bot to negotiate not only on price, but also on additional conditions such as warranty, cancellation options, or bundle discounts – consistently obtained better deals. Others, often without realising it, ended up with less favourable terms.

A bot is only as smart as the instructions it receives. This creates a new type of inequality, not between those who have access to a particular bot and those who do not, but between those who know how to use it effectively and those who do not.

Machine fluency: the new digital divide

According to the researchers, differences in negotiation outcomes stem directly from the way people communicate with the bot. Someone who writes, 'Find the cheapest flight to Barcelona', may obtain a different outcome to someone who provides more strategic instructions, such as: 'Negotiate a flight under €200, including checked luggage and flexible cancellation, and request an upgrade if the price is less than 10% higher.'

In theory, AI shopping assistants could therefore bypass existing shopping platforms and offer users better deals. However, whether this promise is realised depends largely on the individual user's skills. This is a dynamic that regularly emerges with other technological innovations as well.

From the notary to the programmer

Consider, for example, the persistent argument for blockchain that a decentralised land registry could help avoid expensive notary fees. In theory, this sounds appealing: notaries – the gatekeepers who carefully safeguard the quality of land registry data – take a sizeable share of the purchase budget. Would a decentralised database not be cheaper?

In such a database, users propose changes that are validated by other users with an interest in maintaining high data quality (who are automatically compensated for doing so). A win-win situation, perhaps.

However, the question remains as to whether the average prospective homebuyer would be able to work with such a blockchain system.
A new intermediary layer would quickly emerge to assist those who are unable to do so themselves. This layer could consist of computer experts who assist buyers. In other words, programmers as Notaries 2.0.

Alternatively, an automated layer could be built on top of the decentralised database. Think of the interfaces used by crypto exchanges, which allow small investors to trade cryptocurrencies without having to manipulate blockchains themselves.

Those who cannot easily work with the new technology risk becoming dependent on these new intermediaries. And that dependence will not come for free.

Shopping in 2030

The same could happen with AI shopping assistants. Those who cannot instruct a bot effectively may become dependent on new intermediaries, such as prompt engineers or specialised 'bot coaches', or perhaps more likely, the large platforms offering their own bots.

Will the world's largest e-commerce site soon launch a 'Premium Negotiation Bot' that secures the best deals for subscribers only? Or will the top online travel agency introduce a “VIP Booking Assistant” that negotiates exclusive discounts in exchange for loyalty points?

Those who know how to use such tools effectively will have an advantage. But will AI-assisted shopping really save us time and money?
A recent study from the University of Hong Kong challenges one expected benefit: people who shop with an AI bot often take longer to make a purchase. 

According to the researchers, this is because the bot prompts users to pay closer attention to the characteristics of specific products. This is interesting because the more specific a user's needs, the more valuable a product that perfectly meets them becomes. And therefore the higher the (theoretical) willingness to pay.

This will undoubtedly be music to the ears of large platform companies. It is no coincidence that they are currently investing enormous sums in AI development.
 
Will consumers ultimately foot the bill?