AI Agents are already moving beyond answering basic questions and completing research tasks, to acting on instructions when certain conditions are met. No longer do they just scan the internet for the most relevant information in response to a prompt, but they have gone one step further and are now being tasked with finding, securing and paying directly for the content, data and services they consume.
That creates a significant opportunity for content delivery networks (CDNs). Sitting between website owners and the AI systems seeking access to their content, CDNs are well placed to help turn today’s largely uncompensated flow of information into a market for paid access. We are moving from AI agents getting information for free, to a world in which publishers permit, block or charge them for every request.
This dynamic raises an important issue that must be considered before we go full steam ahead. As the internet becomes monetised in this way, it is vital that it does not detrimentally impact the speed at which AI agents provide results. As soon as a financial value comes into play, speed versus security comes into the spotlight.
Disaster recovery is what makes innovation dependable, scalable and secure. If CDNs and AI agents deployed by businesses are going to become participants in a machine-to-machine economy, recovery must be built into wallets from the start. Otherwise, we risk creating an internet that can move value autonomously at unprecedented speed but has no reliable way to get it back when something goes wrong.
From an open web to a transactional web
Under this new model, an AI agent searching for information could encounter a payment request before it receives access to a webpage. It automatically assesses the price against its allocated budget, authorises the charge and continues its task without any human intervention.
Individually, these transactions may be tiny. Collectively, they could create a substantial new revenue stream for CDNs and data owners, especially when you consider the rate at which the use of agentic AI is growing – Chat GPT now sees around 2.5 billion prompts a day, for example.
However, traditional card and payment infrastructure was not built to process huge volumes of micro transactions like these. Fixed fees can make micropayments uneconomic, while the need for human authentication introduces friction to the chain that an autonomous system cannot easily navigate. Stablecoins like USDC, on the other hand, offer an obvious alternative as they can easily be used for micro amounts, and importantly, they settle quickly.
As such, it makes logical sense that digital assets will become integral to the success of the agentic internet. To that end, in order for any business or user to have digital assets, they must have a digital wallet. Enter embedded wallet providers.
They will enable users and services to custody their own funds without having to build their own wallet infrastructure. One wallet may hold the budget delegated for use by the AI agent, for example, while another may receive revenue on behalf of the CDN or content owner. These wallets are invisible by nature but will facilitate these micro transactions.
Learning from prior use cases, adapting accordingly
Gaming offers a useful preview, having long relied on high-frequency, micro transactions for virtual upgrades and rewards within a game. Digital assets extended that model by making value portable across borders and available around the clock, proving their value in the ability to support transactions for which conventional rails don’t make any sense.
But in a world that depends on access to digital asset wallets that users don’t even know exist, there must be thought given to a situation in which that access fails. A failed payment in a game might interrupt one player’s purchase, but a failure at the infrastructure layer of the agentic web could affect tens of millions of automated tasks simultaneously. What amplifies this risk further is the fact that the key infrastructure behind these wallets is concentrated among a very small number of providers.
It essentially means that each provider has an incredibly large share of the market and will likely be underpinning millions of agent wallets. In turn, an outage or catastrophic event at just one of these providers would affect millions of wallets simultaneously, across every service built on it.
What would this look like in reality? Well, if a CDN’s receiving wallet became inaccessible, paid content requests could fail across every service depending on it. At the same time, if the party or business that has delegated budget or assets to an AI agent via an embedded wallet lost access to that wallet because the infrastructure behind it went down, it not only loses access to the funds within it, but the agent also stops working.
To illustrate this even further, think back to those 2.5 billion daily Chat GPT prompts. If every one of those prompts required a fee, and the infrastructure became unavailable, the searches stop and the agentic internet grinds to a halt. Just look at the chaos that has ensued in the past when traditional payments rails have gone down.
When Visa Europe suffered a partial outage in June 2018, the disruption lasted around ten hours and caused 5.2 million transactions to fail across Europe, including 2.4 million in the UK alone. Consumers found themselves unable to pay while commuting, shopping and visiting restaurants and pubs. The incident was serious enough for the Bank of England to conclude that it had caused widespread disruption.
In an agentic world, the impact will be even greater. Rather than individual customers encountering a declined card, the failure of one embedded wallet provider could simultaneously prevent millions of autonomous agents from paying for the information, data and services they need.
Recovery infrastructure is vital
The key to making this new revenue stream practical and scalable is therefore addressing this potential risk head-on, ensuring that the legitimate owner can regain control when access to the wallet is lost.
As of now, controls are in place to ensure that agentic wallets are functioning only according to the permissions granted to them, but there are no protocols in place protecting the owner from infrastructure failure. This is because recovery is not dependent on agents or their permissions. If an embedded wallet fails, no amount of control or regulation of the agents themselves will restore user access. To stop the impact for users at agent level, there must be recovery measures in place at wallet level.
In conventional finance, businesses plan for these events through account recovery, continuity arrangements and tested operating procedures. In digital assets, loss of access can be permanent if no alternative route has been designed and implemented in advance.
Recovery should provide a separate, governed route back to the assets, activated only when defined conditions are met. Depending on the wallet, that could include encrypted recovery material, multiple approvals, independent identity checks, auditable intervention and regular testing. The crucial point is that this cannot be improvised after failure. Once access is lost, it is too late to create a safe recovery path.
The path forward
The industry should adopt a simple principle: no embedded wallet provider should issue a wallet to an autonomous agent, CDN or the users funding them without a defined and tested route back to the assets inside it.
The agentic internet will not scale on payment capability alone. Businesses must trust that transactions can continue, that control can be restored and that assets will not remain trapped because one component fails.
Disaster recovery is what makes innovation dependable, scalable and secure. If CDNs and AI agents deployed by businesses are going to become participants in a machine-to-machine economy, recovery must be built into wallets from the start. Otherwise, we risk creating an internet that can move value autonomously at unprecedented speed but has no reliable way to get it back when something goes wrong.
