In a single week this month, Natural closed a $30 million Series A to build transaction rails for AI agents, positioning itself as a Stripe for agent payments, and Hush Security raised a $30 million Series A to secure what it calls the non-human workforce, taking its total to $41 million.
Two rounds do not make a trend. What makes this worth attention is that they address opposite ends of the same assumption: that software will shortly be initiating transactions and holding credentials on its own behalf, at a volume that breaks the systems built for humans doing those things.
Why the existing rails do not stretch
Payment infrastructure assumes a person authorising a purchase, with the fraud, consent, and dispute mechanisms built around that. An agent buying something on a budget it was given is a different event, and the questions it raises have no established answers. Who authorised it. What limit applied. What happens when the agent was mistaken rather than the merchant fraudulent.
Identity infrastructure has the same shape of problem. Access control was designed for employees, with joiners, movers, and leavers. An agent that exists for four hours, needs credentials to six systems, and is spawned by another agent does not fit that model at all, which is why a security company can raise on it as a category rather than a feature.
Meanwhile the assistants themselves crossed a threshold: the Gemini app passed 1 billion users, joining ChatGPT at that mark. The population of software that could plausibly act on someone's behalf is now enormous.
The commercial consequence for a small product
“A funded rails company can build the mechanism for an agent to spend money. It cannot know that in your sector, an order above a certain value needs a second signature.”
If agents increasingly do the finding, comparing, and in time the purchasing, then how legible your product is to a machine becomes a commercial property rather than a technical detail.
That is not a call to build an integration for something that does not exist yet. It is narrower than that. It means the facts about your product need to be stated somewhere unambiguous and current: what it does, who it is for, what it costs, what it will not do. An agent comparing options cannot infer any of that from a page designed to create a feeling.
A funded rails company can build the mechanism for an agent to spend money.
Most small companies fail this test for a boring reason. Their clearest public description of themselves is out of date, so the confident answer a machine gives about them is wrong in a way no human customer would have been misled by.
Turn what you know into what you own.
Vibepreneur builds structured ventures from professional expertise, with positioning, launch assets, and growth systems included.
Join the WaitlistWhere the actual opportunity sits
Not in payments or identity. Those are infrastructure categories, they are being funded properly, and they will be won by companies who do only that.
The opening for a domain expert is upstream: the rules that govern what an agent is allowed to do in a specific industry. A funded rails company can build the mechanism for an agent to spend money. It cannot know that in your sector, an order above a certain value needs a second signature, or that a particular class of purchase requires a compliance check first, or which exceptions are routine and which are incidents.
Those rules are not written down anywhere. They live in the heads of people who have run the process, and they differ by industry in ways that make a general product useless. That gap is the same one that has always favoured domain knowledge, arriving now in a new form.
What is reasonable to do this month
Very little, deliberately. This is early, and building for it now means building against a specification nobody has published.
What is worth doing is cheap: make sure the public statement of what your product does is accurate and current, and write down the rules your industry applies to the transaction you are closest to. The first protects you from being described wrongly. The second is the raw material for a product, if this continues in the direction two funding rounds suggest it might.
See B2B buyers researching with AI first for the discovery half of this shift, and audit trails are a feature for why the record of what happened is worth selling.