Toward autonomy: D2D payments in the IoT

Let’s begin with a few numbers: a leading market researcher forecasts that Internet of Things (IoT) connections worldwide will surpass 38 billion by 2030. The enterprise segment will account for more than 60% of those.¹ In fact, it predicts a CAGR of 20% by 2030 in smart manufacturing alone.

The IoT is clearly growing and touching every facet of our lives. However, the full value of the IoT will be progressively unlocked as it becomes more autonomous. For the purposes of this discussion, autonomy is about performing business transactions such as the initiating and execution of payments, enabling non-recurring, instant payments between two digital entities, or devices, that are unknown to each other.

What would a world of autonomous device-to-device (D2D) transactions look like, and what would it take to turn it into a reality that businesses find value in? To understand that, let’s take a quick look at where we stand when it comes to making payments in the IoT.

Broadly speaking, there are four generally accepted levels of autonomy within payments in the IoT.

1. Informational, where a device detects a need for a transaction and informs another entity (a human, in this case), but doesn’t initiate a transaction
2. Permissioned, where a device detects a need and moves a request up the chain, and is then authorized by a human to initiate a transaction
3. Conditional, where a device independently triggers a transaction, including payment, but only within a pre-defined rule/parameter. An example would be a device ordering re-supply once it detects a certain level of usage. (Naturally, it would need to have access to a means of payment.)
4. Full autonomy would include a device initiating payment after identifying a need and finding a vendor, helped by adaptive behaviors based on learning and context. An example would be an AI-enabled IoT device managing its own maintenance, including ordering spare parts as needed, with a payment means at its command.

Levels 1 and 2 are already at hand. These are the interactions most people think about when asked to describe payments and other business transactions within the IoT. However, level 3 is not too far in the future – indeed, some use cases that are flirting with it already exist. Level 4 is still at the conception stage. There is currently no scenario where a human hand isn’t required, at least in part.

Research into these higher levels of autonomy continues apace, however, driven by business interest in their applications in the near and long term. Let us consider why enterprises are working toward such an IoT.   

A major factor driving the IoT toward autonomy is convenience. Multiple manual processes – such as selection of vendors, contracting, and settlement – can be eliminated if the device in question identifies the need and initiates fulfillment by itself. An obvious example is a device that takes care of its own re-supply. Reliability is also attractive: empowering a device to perform certain transactions on its own makes it more self-supporting, with less downtime. This has clear applications in the fields of predictive (or even preventive) maintenance. Operational and cost efficiencies would also be substantial, as autonomous D2D transactions enable other processes in the ecosystem to achieve automation.

Of course, for all of this to work, there needs to be a clearly outlined process that increasingly sophisticated devices can internalize and learn from. In effect, this process has to become more human.

Most people wouldn’t recognize all the steps of the transaction process they go through before they leave a supermarket with their groceries. You go to the supermarket because you identify a need; perhaps you need oranges. You know that the supermarket has oranges currently. Once there, you decide whether the price is worth it. You can choose to walk away, or you could think, it is winter, you need Vitamin C, and anyway you’re craving an orange. Perhaps you live in a place where bargaining is OK, and you negotiate what seems like a fair price. You make up your mind and hold an orange to your nose for the all-important sniff test. If it makes you happy, you pay and walk away.

All these discrete steps are so intrinsic to what we do that we pay them no heed. However, a device must be programmed – or learn – to adhere to these steps in order for a transaction to be successfully executed. 

Broadly, such a device would:

1. Identify and assess a demand
2. Find a provider (another device, say) that can fulfill said demand
3. Determine the price of the solution and, if necessary, negotiate a contract
4. Verify whether the good or service delivered meets the demand
5. Once verified, make final settlement, i.e., pay

The more closely a device mirrors human behavior, the more autonomous it can be. 

Interestingly, settlement or payment is not really an issue. The technology required to secure these transactions exists and is in operation. The outstanding point of research currently is verification. How can a device know that what has been delivered is the good or service it ordered, once issues of quality arise, as opposed to a purely quantitative measurement? 

In other words, how will a machine know that the “orange” it ordered actually tastes sweet? And if it were to find that the orange is sub-par, whom would it turn to for redress?

It is clear that trust is integral to this process. Humans establish and live within networks of trust. The grocery mentioned above, where you expect to find fresh fruit at a price you’re comfortable with, is one example. If your fruit turns out sour, you can go back and complain, or even stop going there (and tell your friends!). 

“You trust somebody because you had a good experience with them,” says Philipp Edler, Business Development Manager, Innovation, G+D. “In the IoT, we need to solve that issue at a technical level. In essence, you need to design a mechanism that enables trust, just as you do in any human interaction.” 

Let’s dig into this a bit. A certain base level of trust is assumed every time humans interact financially. Devices must build this trust. Again, this mirrors human behavior.

Firstly, each device has to be sure the other “exists,” beyond an IP address. They must be sure the other one can deliver its part of the transaction, which is a good or service on the one hand, and payment on the other. Security of the transaction must be assured, without tampering or modification on either side. Finally, there must be a regulatory framework that governs the transaction, including a resolution mechanism in case of a dispute.

Establishing trust is part and parcel of the issue of interoperability, which is equally fundamental to payments in the IoT.

By its very nature, the IoT lies at the intersection of connectivity, identity management, and payment. Devices must communicate across networks that follow different protocols and exist subject to the laws of various countries. They must be readily identifiable and secured against tampering and fraud. The payment transactions they execute must inspire trust. 

“The most important issues are security and interoperability, when it comes to payments in the IoT. Those are the questions the industry must answer. Acceptance and adoption will follow once those issues have been addressed,” says Magdalena Dellinger, Technology and Innovation Manager, Corporate Technology Office, G+D.

As devices become more autonomous, the frameworks that govern their ecosystems have to be both sufficiently robust and yet flexible enough to drive growth and innovation. Recent developments in standardization in connectivity across the IoT are an encouraging sign in this regard. Interoperability within the IoT is something all stakeholders need to be involved in. In this context, payments is only one piece of the puzzle.

Artificial intelligence has opened up avenues of development in the IoT that were inconceivable even a few years ago. This is true of payments and other business transactions in the IoT. Machines that can learn from their own experiences have changed the paradigm of what is possible. This includes the interesting field of verification. 

If AI is successful in moving the needle toward an ever-closer mirroring of human behavior among machines, the technology to enable level 4 autonomy will be available at some point. Whether we ever see it – and we might not, due to regulatory or policy interventions – in the future of IoT is another question.