The agent does the shopping, then stalls at checkout
A startup called Natural has raised $30 million for infrastructure that lets AI agents send and receive money on their own. The round was led by Forerunner Ventures, bringing total funding to $40 million. The problem is simple: today's payment rails were designed around human authorization. This issue covers the July 20 agenda and was published on July 25, 2026 to complete the archive.
A payment layer for agents raises money
Natural positions itself as an orchestration layer where AI agents can move and hold funds, make payments, collect money and transact with both humans and other agents. The company announced a $30 million Series A led by Forerunner Ventures, taking total funding to $40 million. Until now the product has run in a closed beta trial.
Why the existing rails fall short
Today's financial infrastructure assumes a person authorizes the transaction. AI agents can increasingly handle work such as finding vendors, comparing prices and arranging deliveries, yet they still need a human to complete the payment. Credit cards and automated bank transfers were never designed for machine-to-machine transactions running at computer speed.
Natural competes on the same ground as players like Stripe here, but says it is pursuing a broader architectural redesign rather than just a faster interface, including how disputed transactions are handled. The company plans to support both dollar-backed stablecoins and traditional bank payments. Founder Kahlil Lalji suggests the number of payments in the world could grow by several orders of magnitude.
Automation breaks on the last metre
An agent finding a product and filling a basket looks impressive, but the chain snaps at payment and the process returns to a person. Making the payment layer machine-friendly means automation can run end to end. It also opens a new risk surface: on whose behalf, up to what limit and under what cancellation terms a payment is made is not a technical detail but a question of commercial trust. That the funding goes toward an architecture covering dispute handling is therefore worth noting.
What it means for businesses in Türkiye
Payment infrastructure in Türkiye sits under its own regulatory framework, so solutions like this take time to become directly usable. Preparation on the merchant side can start today, though. What matters in the near term is that your site and product data are readable by an agent: price, stock, delivery time and return terms published in a way a machine can parse. An agent will not recommend the product with missing information, it will recommend the one with clear terms. On the payment side, two questions are worth settling early: how will you verify that a transaction genuinely reflects the customer's intent, and how will the process work when an automated purchase is returned? In short, open the shop window to machines first, and you will be ready when the payment rails mature.
The UNALSOFT take
Our reading is this: the first winner in agentic commerce will not be whoever builds the slickest interface, it will be whoever has the cleanest data. An agent does not haggle or guess; it decides on what it reads. That is why we build our e-commerce panel work around price, stock and delivery information flowing accurately from a single source. How the payment rails evolve is for time to tell, but fixing product data today pays off in every scenario.
Sources
TechCrunch, "Natural raises $30M to reinvent payments for AI agents"
Is your shop window open to machines?
Let's check together whether an AI agent can read your product data correctly.