Published on the 12/10/2023 | Written by Heather Wright

Billions of new buyers are heading our way, but these machine customers won’t be wooed by two for one deals or celebrity endorsements, and you can’t take them out to lunch or the football to win the deal. 

Instead, Gartner’s Don Scheibenreif says businesses need to prepare for a more programatic, rather than relationship-based, approach to selling in order to get their products in customers’ baskets. But it’s a change that Scheibenreif, who is a Gartner distinguished VP and co-author with Mark Raskino of the book When Machines Become Customers, says is ‘another growth factor for organisations that can be as big as digital commerce’.

“I can’t take an algorithm out to dinner.”

Strictly speaking, Gartner defines machine customers, or custobots, as ‘a non-human economic actor that obtains goods or services in exchange for payment.’

In simpler terms, Scheibenreif compares custobots to the likes of HP’s Instant Ink service, a subscription model that enables printers connected to the service to be monitored and replacement ink cartridges purchased automatically and sent out before the ink runs out. 

“When we think about what we do as human customers, we do five things: We obviously buy things, but we also receive messages from people wanting to sell to us. We ask for service when something is wrong. We negotiate and then we tell other people about our experience – things like user reviews,” he says.

“We’ve been able to find examples of a machine that can do one or more of those activities. Not all five but at least one or more.”

Scheibenreif says there are already 15 billion devices globally that have the capability to become customers – a figure that he expects to swell to 18 billion in a few years.

“So even if a fraction of those devices take on the behaviour of human customers, that represents a lot of money.”

According to Gartner, CEOs believe that, come 2030, up to 20 percent of their companies’ revenue will come from machine customers, which will be involved in a wide range of consumer and business purchases. Gartner itself has forecast that by 2030, machine customers will be directly involved in, or have influence over, trillions of dollars of purchases. 

As the numbers – and capabilities – of custobots increases, organisations will need to reconsider how they sell to those customers, and figure out what the algorithm is to get into a customer’s shopping basket.

Machine customers will operate based on rules and logic, and will be more likely than human customers to make efficient purchases. That takes the form of smaller product sizes, minimising waste (you might buy five apples a week, but your refrigerator knows you only actually eat two a week and the rest are thrown), substitutable products that might be geographically closer with lower shipping costs, or recommended value-added products that might be more expensive in the short term, but cheaper in the long run.

“Say I’m a manufacturer selling widgets to a retail store. What if that retail store decided to outsource its negotiation function to a machine? I can’t take an algorithm out to dinner. I can’t take it to a football game. But what does it need? It needs information.”

Scheibenreif says the rise of the custobot – something he’s adamant is already starting to happen – changes the entire selling process.

Stripped of the option of using FOMO (fear of missing out) or other emotional drivers, including ‘cool’ celebrity endorsements, to drive sales, it becomes less about relationship-based selling and more about programatic selling –  or ‘what is the combination of factors that allows me to get my product into your basket?’

In Italy, what has been billed as ‘the world’s first machine customer commerce platform’ has already launched, providing factory equipment to request replenishment supplies from a central marketplace – as opposed to just the original vendor.

Scheibenreif also cites the example of OpenAI co-founder Greg Brockman using ChatGPT to come up with a dinner party menu which was then sent to Dall-E, OpenAI’s text-to-image offering, to provide an image for the invitation, which ChatGPT then sent and asked for an ingredient list, which in turn was sent to the Instacart grocery delivery service. All done within a matter of minutes.

“The use of ChatGPT is taking on some of the behaviours of machine customers,” he notes. “It’s that type of stuff that is going to be interesting because those capabilities extend, or make it easier, for a machine to be a customer.”

Initially, custobots will be rules-based, with users setting the variables. However, Scheibenreif says over time there will be more discretion for the machine. 

“I don’t see that happening right now but generative AI might change that because it is not necessarily rules bound – it’s bound by the information it is trained with. So if it’s trained on all the possible pasta brands out there, and I say I want this one, then I imagine the generative AI may say ‘this brand is pretty similar to this brand. It wasn’t on your list but it might be an acceptable substitute.’

He believes a shakeout is coming for categories where products are largely the same, despite claiming to be different.

But there are ‘a whole bunch of things that still have to be worked through’ he admits.

One of those is the issue of trust and knowing the machine you’re interacting with is real, and not a scam.

Scheibenreif says a large credit card provider has talked about wanting to be a trust broker. 

“So there’s an element of trust and verification that has to be there. Identity access management already exists today, and there are some trust certifications out there, but nothing for machines.”

Despite that, Scheibenreif is calling on businesses to to start getting ready for the new customers. 

“The rise of the custobots will change the fortunes of organisations that take this trend seriously,” he and Raskino say.