Your Next Customer Might Not Be Human. Get Ready Now.

On July 20, a one-year-old startup called Natural announced a $30 million Series A to rebuild payment infrastructure for AI agents. Forerunner's Kirsten Green led the round, which brings total funding to $40 million.

The founder story matters here. CEO Kahlil Lalji already built and sold one fintech (Ivella, a YC-backed banking product for couples, acquired by Earnin in 2023). He told TechCrunch he'd been burned by fintech and wanted to avoid the sector entirely. Then he couldn't stay away: "It just feels obvious that agentic payments are going to be structurally the most important problem in the space."

And he's not alone in that read. Stripe is racing to redesign its own rails for agents. Skyfire Systems is building an agent payment backbone on stablecoins. When a repeat founder who swore off fintech, the biggest incumbent in payments, and a cluster of venture-backed challengers all converge on the same problem in the same summer, that's a signal worth reading.

Because its owner told it to. Agents in mid-2026 already handle real procurement tasks: finding vendors who can move freight, comparing prices, messaging the vendor, and scheduling delivery. The one step where they still stall is the payment, which is exactly the gap Natural wants to close.

Zoom out and the buyer side gets more interesting. Jack Dorsey launched Buzz on July 21, a group chat platform built for teams and their AI agents, positioned squarely against Slack. When agents sit inside a company's chat as coworkers, they inherit errands. Renew the analytics subscription. Find a cheaper transcription API. Book the contractor.

So the question stops being hypothetical. If an agent showed up at your pricing page today with budget and authority, could it complete a purchase? For most startups the answer is no. There's a demo request form, a sales call, a CAPTCHA, and an email verification loop standing in the way. Every one of those is a wall to a machine buyer.

Almost everything, because your checkout was designed to slow down fraud committed by humans, not to serve legitimate software.

Traditional rails like credit cards and ACH assume a human authorizes each transaction. That assumption shows up in small ways all over your funnel: SMS verification codes, "click all the squares with traffic lights," billing forms that require a browser, terms of service written only in legalese prose. An agent trying to buy hits these speed bumps and either fails or hands the task back to its owner, who may just pick the competitor that didn't make them intervene.

Here's the thing though: you don't need to solve authorization yourself. Natural, Skyfire, and Stripe are fighting to be the layer that handles agent identity, spend limits, and disputes. Your job is narrower. Make sure that when those rails mature, your product is the easy thing to buy on them, not the one that still demands a phone call.

Start with information, because before an agent buys, it researches. Agents choose vendors the way answer engines choose citations: they pull whichever option is clearly documented, clearly priced, and machine-parseable.

A few concrete moves. Publish your pricing. Hidden "contact us" pricing is invisible to a software buyer running a comparison. Keep a clean, public docs site; agents read documentation better than they read marketing pages. Add structured data to product and pricing pages so parsers don't have to guess. Write a plain-language summary of your terms next to the legal version.

Then look at the transaction itself. Can someone go from landing page to active account without talking to a human? Self-serve signup, API access, and usage-based entry tiers are the agent-friendly path. None of this is wasted if agentic commerce arrives slower than expected. Every step also makes you easier to buy for impatient humans, and there are plenty of those.

Maybe not today, but you should know which direction the pressure pushes. Agents don't sit through 30-minute demos or annual contract negotiations. They compare, decide, and transact in seconds, which favors usage-based and per-transaction pricing over seats and sales cycles.

The scale argument is what makes this worth planning for. Lalji's bet is that if transactions happen at computer speed instead of human speed, "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." You don't have to believe the full version of that claim. Even a 10x rise in machine-initiated purchases would reward products that can be bought in small, frequent, automated increments.

A practical middle path: keep your current pricing for humans, but sketch what a metered, API-first tier would look like. Know your unit costs well enough that you could ship that tier in a quarter if the market moves.

You probably shouldn't build payment rails; that fight already has $40 million challengers and Stripe in it. But infrastructure shifts create picks-and-shovels openings one layer up, and those are very much small-team territory.

Some examples worth stealing. Agent-friendly storefront tooling for the millions of businesses whose checkout can't serve a machine buyer. Middleware that translates existing product catalogs into structured, agent-readable formats. Monitoring that tells a vendor when agents are visiting, failing, and abandoning. Dispute and refund workflows for purchases no human remembers making, which is a genuine mess waiting for a product.

The pattern from past shifts holds: when the big players build highways, the durable small businesses sell what the traffic needs. Mobile got app stores, then a thousand small companies got rich on tooling. Agentic commerce looks set to repeat that shape, and in July 2026 it's still early enough that nobody owns the second layer.

Real, and worth naming. Natural has been operating in beta. Standards for agent identity and authorization aren't settled. Stripe could absorb the whole category the way it absorbed billing. If you rebuild your business around agent buyers this quarter, you're betting on a timeline nobody can promise.

So hedge. The goal is agent-ready, not agent-only. Everything in this playbook (public pricing, clean docs, self-serve signup, structured data) pays off with human customers immediately, which means your downside is roughly zero even if agentic commerce takes three more years to matter.

This is a classic two-scenario planning exercise, and it's worth doing on paper rather than in your head. Sketch scenario one, where agent-driven purchases stay under 1% of your market through 2028, and scenario two, where they hit 15%. You can map this in a spreadsheet, Notion, or a planning tool like Foundra that walks first-time founders through go-to-market scenarios step by step. What you're looking for is the moves that are right in both worlds. Do those first.

A short list, in order of effort.

First, run the agent test on your own funnel. Ask an AI assistant to research your product category and watch what it finds and recommends. If you're absent or wrong in that answer, fix the information problem before anything else.

Second, publish real pricing and a plain-language terms summary. Third, audit signup: count every step between landing page and value, and cut the ones that exist for no defensible reason.

Fourth, if you sell software, expose an API for your core action, even a minimal one. Fifth, set a calendar reminder to check the space quarterly. Watch Natural's launch out of beta, Stripe's agent announcements, and whether platforms like Buzz turn agents into daily coworkers at normal companies.

That's a few weeks of work total, most of which improves human conversion right away. Cheap insurance against a shift that three well-funded teams are betting will be structural.

What is agentic commerce? Purchases researched, negotiated, or completed by AI agents acting for a person or business rather than by the person directly. Payments are the current bottleneck, which is what Natural, Skyfire, and Stripe are working on.

Do AI agents actually buy things in 2026? They handle vendor research, comparison, and ordering steps today, but most payments still need a human in the loop. The infrastructure to remove that step is what just attracted fresh venture money.

Should I add "agent support" to my roadmap? Add agent readiness: public pricing, docs, structured data, self-serve signup, an API. Skip building custom agent integrations until standards settle.

Will agents kill sales-led startups? No. High-trust, high-price sales will stay human for a long time. The pressure lands first on low-consideration, comparable purchases like tools, subscriptions, and commodity services.

Who wins if this doesn't happen? You still do, mostly. Nearly every agent-readiness step also removes friction for human buyers, so the work is low-regret either way.

1. Natural raises $30M to reinvent payments for AI agents and take on Stripe (TechCrunch, Jul 20, 2026)
2. Jack Dorsey is taking on Slack with Buzz, a group chat platform for teams and their AI agents (TechCrunch, Jul 21, 2026)
3. Natural (company site)
4. Almost 90 new unicorns have been minted so far this year (TechCrunch, Jul 5, 2026)