Natural raised $30M to build the payments layer for autonomous AI agents

The Machine Is the Customer Now: Natural's $30M Bet on Payments for Things That Aren't Human

Every payment system in use today, from Stripe checkout to ACH transfers, was designed around a single assumption: a human is authorizing the transaction. That human has a credit score, a government ID, and a phone number a bank can call when something looks suspicious. Remove the human and the supporting architecture, identity verification, fraud detection, dispute resolution, becomes scaffolding with no building underneath it.

A startup called Natural just raised $30 million to rebuild financial infrastructure for exactly that world. The Series A was led by Kirsten Green at venture firm Forerunner, bringing Natural's total funding to more than $40 million 1. The company positions itself as an "agent orchestration layer" that lets AI agents move and store funds autonomously 2.

AI agents are already identifying vendors, comparing prices, and negotiating deliveries 2. The bottleneck is the moment money needs to move. Today the agent hands off to a human with a credit card. Natural's founders see that handoff as a problem to be eliminated, not a permanent feature of commerce.

The structural gap runs deeper than convenience. Consider what happens when a payments system tries to onboard an AI agent as a customer. KYC regulations assume a verifiable natural person or legal entity. Fraud scoring models are trained on human behavioral patterns: typing cadence, IP addresses, spending velocity. Chargeback rights depend on a cardholder who can call the bank and say "I didn't authorize this." None of these guardrails map onto a language model calling an API at three in the morning to buy compute time from a vendor it found through a web search.

Natural's answer is to build the entire stack from scratch. The company is developing 13 products, with six now generally available: FDIC-insured wallets for agents (banking services provided by Column N.A.), one-way vault accounts that let agents receive but never withdraw funds, and payment primitives for sending, requesting, and transferring money between agents, businesses, and consumers 1. Products launching in coming months include debit and charge cards for agents, and Q4 brings per-API-call billing, credit lines for agents, and success-based billing 1.

CEO Kahlil Lalji previously co-founded Ivella, a YC-backed banking and financial product for couples that was acquired by Earnin in 2023 2. He says the company needed to own the full stack rather than build atop existing rails. Natural built its own ledgering, money movement, multi-bank settlement, multi-currency handling, fraud compliance, and what the company calls "agent identity observability" 1. That last phrase signals the core problem: verifying and monitoring what is spending money, not who.

The incumbents are not idle. Stripe is also racing to redesign payment rails for AI agents 2. Other startups are attacking the same gap with different tools: DCVC-backed Skyfire System is building agent payments on USD-backed stablecoins 2. Natural plans to support both stablecoins and traditional bank payments, betting that breadth across rails will matter more than loyalty to any single one 2.

But the hardest questions are not technical. They are legal and structural. When an autonomous agent overspends, who bears the loss: the developer who built it, the business that deployed it, the infrastructure provider, or some category of entity that does not exist in financial regulation yet? Dispute resolution today depends on a human cardholder picking up the phone. No framework exists for an agent that autonomously disputes a charge it calculates was mispriced. And fraud liability, which flows through a well-understood chain from issuing bank to merchant to card network, has no established path when the cardholder is software.

Lalji's thesis on why this matters now is about transaction volume, not novelty. "The number of payments that may occur in the world may be two or three or four orders of magnitude greater than the number of payments that exist today" 2. If payments happen at computer speed, the infrastructure needs to handle thousands of times more of them, with no human in the loop to catch errors.

Natural has a team of 17, has attracted engineers from Stripe, Ramp, and Square, and was founded 342 days ago 12. The Series A was raised when the company was 193 days old 1. That speed is the bet. The incumbents have more money and more customers, but they also have payment systems optimized for humans. Natural is starting from zero, and zero comes with fewer assumptions.