Computer algorithms are taking over the business of buying and selling products online.
Already a subscriber?In mid-2020, Bonds, a well-known Australian fashion retailer, released a line of cotton face masks that had been treated with an antiviral finish. Back then, most people assumed wearing a cotton mask provided good protection against COVID-19, but masks were in short supply around the country. Unsurprisingly, they sold out in an instant. The company had to make a rush order of another 4 million to keep up with demand.
The algorithm I used to buy my face masks was simple. There are plenty of algorithms available on the internet that do such basic tasks completely free of charge – and if you like them, you can hire them for more complex, paid jobs in the future. More capable algorithmic agents can perform every single step of a transaction. The only things that slows them down are captchas – but more on those later.The algorithms that buy and sell products and invest money on our behalf deserve a name.
There are many ways digital minions can wreak havoc beyond disrupting the markets. We all have to be aware that things can go wrong, but I tend to be an optimist and prefer to consider opportunities and the positive impact algorithms can have.When I’m not busy thinking about algorithms, I love to go on trail runs. A couple of days before writing the first draft of this chapter, I went out for my usual run. It was a winter evening, and the forest was already dark.
This task was more urgent, and more important, than any others, including getting up. I managed to press the button within the grace period the watch had given me. Thank you for taking care of me, dear sports-watch algorithm, I thought, but I don’t need any assistance this time. Chavous had to convert the GPS location to a point on a map and then provide that location to the rescuers. Within 12 minutes of the watch asking for help, the rescue team located the owner in a shrubby area near his home. The dispatchers said the watch saved the man’s life.I am quite fond of these quiet heroes: algorithms that churn away doing useful work behind the scenes.
Of course, if a business can replicate its products or services efficiently and virtually without expense – that is, if additional units can be produced at near zero cost – it’s good for profits. The IT company HP recently introduced a new service in Australia, called HP Instant Ink. Now that more people work from home, office-printer usage has dropped. In June 2022, the chief executive of HP said it was expecting a 20 per cent decrease in office-based printing compared with its pre-pandemic estimates.
The ability of HP’s digital minion to anticipate the need for a new ink cartridge and order it before the previous one runs dry makes it a proactive algorithm. Proactive digital minions generally take on typical consumer tasks: they can analyse needs, choose from different options and then execute transactions, sometimes waiting for human consent first. Some are more sophisticated, and some are more limited.
Hard-coded digital minions are a simple strategy, and they work, but some customers might feel trapped by the lack of options they provide, and competing service providers won’t be happy either. Sooner or later, regulators will force manufacturers to let customers choose their own provider. We do need to be mindful that it is possible for a proactive digital minion to breach the privacy of individuals simply by acting upon its predictions. In a well-publicised case, the US retailer Target used sales data to predict which of its customers were pregnant, and their stage of pregnancy. This information is a retailer’s dream: suddenly, it can offer so many targeted products to the customer.
But the “no limits” part turned out to be misleading: it was impossible to buy PlayStation 5 on its launch day. Whenever it showed up in online stores, the console sold out within seconds. Some online store websites crashed immediately after the launch, and others experienced strange glitches: customers reported that the console would appear to be in stock until they clicked the “add to cart” button, at which point the website informed them the product was not available.
This phenomenon of digital minions purchasing goods and services on behalf of humans is so common it needs a label. I call it B2A2C – an acronym that should be recognisable to business school graduates, but which – I’ll be the first to admit it – might be a bit too much for everyone else. It describes a new type of business relationship that has emerged in addition to business-to-business and business-to-customer relationships. It stands for “business to algorithm to customer”.
Mediation occurs when software agents are created by a third party to help bring businesses and customers together. Almost like personal shoppers, these digital minions search for a product the customer needs and then order it. There are many scenarios in which such algorithms would make sense: shopping for the best insurance deals or buying coffee beans from local sellers.
He is open about how things sometimes go wrong. His biggest failure? That one time he spent $3000 to hire a large number of bots to “cart” a hundred pairs of sneakers that “dropped” that day. That’s a cost of $30 per pair, which would have been easily absorbed by his margins. But he only received three pairs of sneakers in the mail, making the cost $1000 per purchased pair, which would be impossible to make back.
In January 2023, DoNotPay’s CEO, Joshua Browder, tweeted that his company was willing to pay a volunteer $1 million to appear before the US Supreme Court at an upcoming trial and argue the case by repeating exactly what the robot lawyer would tell them to say. Browder suggested that the robot could communicate with the person in real time via AirPod headphones. “We have upcoming cases in municipal court next month,” Browder noted, “but the haters will say ‘traffic court is too simple for GPT’.
Is the world of B2A2C only within the reach of tech-savvy users such as Nova or Silicon Valley businesses? Thankfully not: that’s where enablement – the third category of the B2A2C approach – comes into play. Just as platforms such as eBay created opportunities for almost anyone to sell products online, there are now platforms that allow businesses to deploy digital minions without building the entire infrastructure from scratch.