For the first time in history, over half of all internet traffic is now attributed to non-humans according to an industry report, and it’s set to rise yet further as agentic AI (AGI) gains ground. AGI is the next step in the evolution of Generative AI and will see the technology become fully independent without the need for human prompts. These autonomous agents will then be able to transact, subscribe, book and buy making them the primary participants in the digital economy.
We’ve already seen the likes of VISA, for instance, announce its intention to support AGI payments. The payments handling company foresees a future where the technology will carry out mundane purchases on our behalf such as for groceries, clothing and plane tickets based upon our previous shopping habits. It has even stated that AGI will be as transformational as ecommerce but the company also rightly acknowledges that the technology faces a major hurdle: the need for human input to access and pay for goods and services.
Today, websites make the assumption that every brand interaction is with a user at the end of a keyboard. This effectively locks out AGI, which then cannot access, pay for, or interact with digital content which is protected by identify and security measures such as logins, anti-bot protections or compliance layers. To overcome this issue, we need to overhaul the online business model by rethinking how data is accessed, consumed and governed via Application Programming Interfaces (APIs).
AI and APIs
Every AI prompt, data request and access attempt involves the use of APIs. They’re the unseen elements that connect and facilitate real-time digital communications and transactions but they can also act as a barrier. APIs can create friction in human transactions but doubly so when it comes to AI which is simply not recognised as a valid entity. So how can businesses ensure they are prepared for the move from a human to a AGI-friendly digital economy?
Critically, AI agents need to be validated at the point of interaction so this will require a new level of secure, programmable access to websites, APIs and applications. Rather than requiring a human to sign-up, verify and login, we need to adapt the process to allow AI agent to present verified credentials and payment methods. This avoids the need for a pre-existing contract to be in place or a human to manually create an account but ensures the transaction is still carried out over a secure peer-to-peer connection.
In practice this works by the AI agent making a service request via API to the website or service provider with a specific token included in the header enabling identification. The service provider then verifies the token which has the agent’s identity and payment information. The final step sees the payment processed, the service delivered and the payment charged to the virtual wallet of the agent which may include various funding sources such as cards, an automated clearing house (ACH), wire or cryptocurrency.
An AI-driven economy
It’s a relatively simple process but one that could have a profound effect. Freed from the constraints of needing a human to complete the process, these AGI agents will be able to act on our behalf leading to greater efficiency. They’ll be part of a global payments and identify network which will lead to AI-driven interactions happening at scale in milliseconds at any time of night or day. As such we can expect it to dramatically reshape the digital economy, leading to new opportunities and revenue streams.
However, it’s worth issuing a word of caution at this point. With over a third of internet traffic currently comprising malicious bots, its imperative that businesses have the means to distinguish between legitimate AGI requests for access versus these bad bots. Therefore, in addition to verifying the agent at the start of the process it’s necessary to observe the context, behaviour, and intent during the AI-driven API interaction.
Machine learning together with behavioural analytics can be used to anticipate the expected actions of these agents and create parameters of behaviour. If the AI then does not operate within those parameters, enforcement measures can be taken such as blocking the AI in the event a transaction appears fraudulent. In this way, security can be applied in way that does not create a barrier to business but ensures the service provider is protected.
Adopting this approach will solve the problem of AI lockout. It will enable businesses to verify and trust agents at the edge and extend the same access privileges to AI as we do humans, emancipating the technology. In doing so it will usher in a new era of trusted automation and build confidence in the capabilities of the technology. It’s only then that we will see AGI fulfil its potential, become the new consumer and allow an AI-driven economy to flourish.
