In July 2024, Delta Air Lines announced plans to expand its use of AI-powered pricing from 3% to 20% of its flights by year-end. The airline partnered with Israeli tech company Fetcherr to set fares based on predictions of "the amount people are willing to pay," according to statements from Delta President Glen Hauenstein to investors. Delta insisted it was not using data about individuals, but the company had already told investors it was working toward a future where "we will have a price that's available on that flight, on that time to you, the individual."

Earlier in 2024, Wendy's triggered a backlash when CEO Kirk Tanner announced plans to test "dynamic pricing and day-part offerings" at its restaurants starting in 2025. News outlets interpreted the announcement as "surge pricing" – the practice of raising prices during peak demand, made familiar by Uber. Within days, #BoycottWendys trended on social media, and rival Burger King launched a "No urge to surge" promotion. Wendy's backtracked, insisting it had never planned to raise prices during busy periods but only to offer discounts during slow times.

Artificial intelligence is broadly reshaping how much consumers pay, for everything from airline tickets and accommodation to clothing and transportation. Dr Sam Kirshner, an Associate Professor in the School of Information Systems and Technology Management at UNSW Business School, explained the distinction between pricing models that have existed for decades and those that have evolved with the advent of artificial intelligence. A/Prof. Kirshner gave the example of pubs offering drinks during quieter periods.

"The key thing there is that the prices are the same for everyone who walks in the door. Doesn't matter where you're from, what your age is, your gender, everyone gets the same price," said A/Prof. Kirshner, who was recently interviewed by Dr Juliet Bourke, Adjunct Professor in the School of Management and Governance at UNSW Business School, for The Business Of, a podcast from UNSW Business School.

Personalised pricing operates differently. Companies have collected vast amounts of information about individuals and use that data to determine what each person might be willing to pay. "In most cases, the characteristics that they're using are also dynamic,” A/Prof. Kirshner explained. “So potentially, where you are in the world, the internet connection that you're using could even just change the price, even with all your other data being constant."

The mathematics of what you pay

A lack of transparency around AI pricing has created a sense that consumers are operating in the dark. "It is just so opaque in terms of what they are actually using to sell me this flight at this price. Is it my cookies or my browsing behaviour, or just the country or city that I'm in, that is determining this price?” asked A/Prof. Kirshner.

He described a case involving a book about the genetics of flies, where two online sellers each used the same strategy: automatically price their copy higher than the competitor's listing. The logic seemed sound – if someone bought from them, they could purchase from the other seller at the lower price and still make a profit on the markup. But when both sellers ran this algorithm simultaneously, the price spiralled upward – eventually reaching a price of more than US$23 million. “The plan was, ‘if we sell it, I'll just buy it from the other person and then sell it at a markup.’

Learn more: How Uber drivers revolt against algorithmic management

“So even just this rudimentary algorithm, which wasn't designed to be anti-competitive in any way, shape or form, just ended up being like that. And now with much more sophisticated AI, there's lots and lots of research that shows, if you just use machine learning to come up with these prices very quickly, these algorithms figure out how to be collusive,” he explained.

The data trail behind your price tag

The information that feeds pricing systems is from different sources. Companies track location data, browsing patterns, purchase history, and the types of devices used. A/Prof. Kirshner gave the example of bike-sharing companies that flooded the streets of Sydney with e-bikes around 2017-2018, and he suggested they may have been collecting travel data as part of their business model. "I'm pretty sure that was just a strategy to figure out where people travel, just to collect that data and to sell that data onwards. And so just that data, when combined with other pieces of information, can give very detailed portraits of people," he said.

The process by which this data translates into prices is complex. Rather than assigning weights to factors like age or postcode, the systems use learning to identify patterns across datasets. A/Prof. Kirshner compared the approach to advertising on networks, where algorithms select content based on correlations within certain amounts of information.

Some consumers have learned to work within these systems to secure better deals. A/Prof. Kirshner described how people who understood the mechanics of online pricing could game the system. These shoppers would add items to their online cart but deliberately wait, sometimes for 30 minutes or even a full day, before completing the purchase. Retailers, eager to close the sale, typically send follow-up emails offering incentives such as free shipping or discounts of 10% or more, based on algorithmic recommendations.

“So, in many ways, the people who are more savvy and have a better understanding of this are probably going to end up with better deals. However, if you're not as literate in using tech and you only know this one website or you only know this one app, then that can be exploited,” said A/Prof. Kirshner.

Australia's position on consumer protection

Australia has maintained stronger consumer protection regulations regarding privacy and data than the United States, but not as stringent as those in the European Union, A/Prof. Kirshner explained. And within these respective regulatory restrictions, markets have generally allowed companies freedom in how they deploy these technologies.

This has created a trade-off between personalisation and privacy and security, depending on where consumers are in the world. “If everything is secure in terms of our data, then you can't use that data to actually target people, right?” A/Prof. Kirshner explained. “So, in this case, yes, we're safer. Things are fairer, but you might not get the convenience and the product recommendations that you want."

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

The convenience that personalisation offers also comes with a cost that extends beyond transactions. When companies possess detailed pictures of individual consumer preferences, they can identify which products each person values most. A/Prof. Kirshner said this kind of knowledge puts pressure on prices, since firms can charge more for items they know consumers want.

However, he noted that some competition could emerge from personalisation. When companies believe they had a chance of securing a sale based on consumer data, they compete more, which can drive prices down. However, A/Prof. Kirshner said the question remains whether Australia will set thresholds for levels of data use in pricing.

The path toward business practice

Dr Barney Tan, a Professor of Information Systems and the Senior Deputy Dean (Impact and Partnerships) at UNSW Business School, pointed to examples of companies that had begun to address transparency concerns.

Microsoft has developed AI principles that include defining areas where the company refuses to apply AI, regardless of the potential profit. Google has introduced model cards that explain how AI models function and what data they use. "Think of them like nutrition labels for AI," Prof. Tan said. "Now these aren't just compliance tools. They're trust-building mechanisms when consumers know what data is being used and what isn't, they feel safer, and that's crucial."

However, some companies are hesitant to invest in certain AI practices, and Prof. Tan said these organisations often wait to see how competitors fare before committing resources. "We see a lot of mid-tier companies that are hesitant to act,” he said. “They're not investing heavily in responsible AI yet. Instead, they're watching the market, waiting to see whether the big players get away with opaque pricing models, or whether they get regulated or shamed into reform."

Market forces and consumer awareness

A/Prof. Kirshner emphasised that businesses need to think beyond profit opportunities when implementing AI pricing, and he warned against companies justifying practices by pointing to competitors that engage in potentially unethical behaviour. "Everybody in the industry is doing it. So, how bad is it if I just do it too? So it's kind of like diffusing their responsibility," A/Prof. Kirshner said. "What competitors are doing is often a bigger driver than just, ‘oh, we can earn more profit if we do this right.’"

He suggested business leaders put themselves in the position of customers facing the pricing systems their companies might create. "If you are a consumer, and you put your data in and you get a certain price, just like putting the mirror on you: How do you feel about this? Do you find it fair?" he asked. While this test seems straightforward, he said examples from technology companies suggest they do not apply this approach.

Learn more: The evolution of dynamic pricing: should AI decide what you pay?

Drawing parallels to fashion, A/Prof. Kirshner suggested that consumer demand could influence how companies approached pricing ethics. The rise of clothing brands showed that markets could develop for products aligned with consumer values. However, the success of fashion retailers like Kmart and Target demonstrated that convenience and cost often outweighed concerns.

"Hopefully there will be a demand for privacy, for fair pricing, whether that's static or just limiting what types of data are used in a more transparent and a more systematically algorithmic way, so that you can actually see how personal characteristics are weighted in order to generate a certain price," A/Prof. Kirshner said. "But I think, ultimately, people's convenience will potentially win out."