AI agents can already write code, build websites, and manage workflows … but they still can’t pay for anything on their own. That bottleneck is about to disappear.
In this episode of TechFirst with John Koetsier, we sit down with Jim Nguyen, former PayPal exec and cofounder/CEO of InFlow, a new AI-native payments platform launching from stealth. InFlow wants to give AI agents the ability to onboard, pay, and get paid inside the flow of work, without redirects, forms, or a human typing in credit card numbers.
We talk about:
• Why payments — not intelligence — are the missing link for AI agents
• How agents become a new kind of customer
• What guardrails and policies keep agents from spending all your money
• Why enterprises will need HR for agents, budgets for agents, and compliance systems for agents
• The future of agent marketplaces, headless ecommerce, and machine-speed commerce
• How InFlow plans to become the PayPal of agentic systems
If AI agents eventually hire, fire, transact, and manage entire workflows, someone has to give them wallets. This episode explores who does it, how it works, and what it means for the economy.
Transcript: PayPal for agents, welcome to the age of agentic commerce
This is a lightly-edited transcript. Refer to the full video for exact quotes.
John Koetsier:
Soon AI agents are going to be employing, billing, and paying each other. Hello and welcome to Tech First. My name is John Koetsier. Everybody’s building agents. We know this, we see this. We’re building sites. They’re spinning up infrastructure, they’re orchestrating workflows, they’re entering data, cleaning up documents.
But let’s be honest, they’re not full-fledged teammates yet, if they ever will be. They can’t hire, they can’t onboard, they can’t pay, they can’t say, “I don’t do this, but I know an agent that does that.” So agentic commerce, agentic automation is a missing link in agentic automation — perhaps not for long.
Here to chat is Jim Nguyen. He’s a former PayPal executive and current CEO of InFlow, which enables AI-native payments. Hello, Jim. How you doing?
Super pumped to have you. It’s a cool topic. It’s a super interesting topic. I get pitches on agents daily — I think that’s accurate. And sometimes it’s more than one. But I haven’t got a pitch on [00:01:00] agents that pay each other, agents that employ each other, onboard each other, all that stuff. It’s a super cool concept and idea. It’s a natural progression, once I’ve seen it, of where we’ve been.
I’m gonna quote from your press release. Your press release says: Every platform shift redefined how money moved. PayPal made payments native to the web. Apple made them native to apps. Now InFlow makes them native to AI agents — agentic payments that are frictionless, trusted, and autonomous. Super interesting statement. Just a big, broad question: why do agents need financial capabilities?
Jim Nguyen:
The missing link is: how does this work become commerce?
So what we enable is the ability for an agent, who’s a new customer, to transact with others. We think of humans as customers — maybe your children are customers of the web and mobile — but this is a new customer that has completely different ways to interact with payment systems.
So we built InFlow with the mindset that it’s a new customer. What does this customer need, and how does this customer interact when they buy and sell and transact with others? That’s what we built it for.
John Koetsier:
Huh. Give some examples. What does this look like?
Jim Nguyen:
As we can see, the AI market is going crazy.
There are AI coding agents, there are a lot of workspaces out there. When a person sits and uses these AI agents or applications, they might say, “Hey, please go build me a website.” The coding application gives you a beautiful website, but that’s where it stops, because the next thing you’re gonna ask is, “Hey, I’d like to be able to host this on a database and get a domain name for it,” right?
So the agents can build the website, but the human needs to step in and create accounts for the database, create an account for the domain name, then pass it back to the agent. That is the friction we’re solving. We’re enabling agents to embed onboarding and payments inside the workflow so they can create a new account on the database, pay for the database, and then provision that database — all within the IDE or workspace the developer is in.
John Koetsier:
That is super cool. As somebody who’s built and published dozens — if not scores — of websites, it’s challenging. Build or buy or commission a theme; if it’s WordPress, get your hosting, upload the theme, customize it, customize your install, add e-commerce capabilities, hook it into your stats package from Google or whatever.
All this is very manual. And you’re talking about taking that and [00:04:00] saying, “Hey, I can vibe-code this website, and now I can hook up all the wires with agent workflows.”
Jim Nguyen:
Exactly, because the agents are smart enough to build websites — they have color coordination, all these things.
We think it’s easy to create a new account. But the issue is not intelligence, it’s the lack of a payments infrastructure to enable this. So that’s critical. When we started this company, of course we were building a payment system.
The way we’ve centered our approach is imagining ourselves as a payments team within our customer’s team. The CEO comes and says, “Hey Jim, we launched a product six months ago, it’s doing great revenue. How can we accelerate the payments? What can you do?” That’s the mental model we’re working with.
Now that my AI coding agent can take payments from a human, can my agent take payments from an agent?
John Koetsier:
Mm-hmm.
Jim Nguyen:
Right? Because if you don’t support the agent, you’re going to lose a customer — a buyer.
John Koetsier:
Mm-hmm. Mm-hmm.
Jim Nguyen:
So now that we’ve identified friction, the two friction points are onboarding — much like what humans do — and configuration of payments, which you must do again.
So we ask: what technologies are available to make this seamless? Our approach is top-down: what is the consumer experience, and what does a company want that experience to be? Then build a payment system around that. That’s our differentiator.
John Koetsier:
Super cool. Two thoughts come to mind. One is the customer experience — it’s going to be amazing and incredible if this works well, because you can provide the full meal deal. Not just one slice of what you need to get up and running or get a workflow operational.
It’s everything, with different companies and different agents brought in to help out. The second thought is the maturation of AI agents. The human concept of an agent predates AI — somebody who’s my representative, has authority, can negotiate, pay for me, act for me. And that hasn’t truly been the case with AI agents yet, correct?
Jim Nguyen:
That is correct. AI agents are super smart — they have the index of the whole web, every book ever written — but the mechanics of interfacing with payments is a different layer.
Just like when you’re building a SaaS company, the SaaS application is different from the payments you’re putting in to monetize it. The intelligence layer is solved. What’s missing is payments — and not just payments like “I have a wallet, I have stablecoins, I have fiat.” That’s table stakes. It’s understanding the customer experience and what’s blocking them. That’s critical, and that’s the top-down way we built InFlow.
It’s in the flow of work.
John Koetsier:
Cool. Is this primarily for businesses? A B2B play? Or is this for consumers? It looks like B2B with a strong consumer addition. Am I right?
Jim Nguyen:
The answer is on a timescale. In the short term, it’s whoever wants to sell and generate more revenue from a new customer base.
Over time, as AI matures, you’ll have agents working with and for each other, paying each other. One of your segments said every person will have a hundred AI agents. Well, are those agents working? And if they are, are they paying and getting paid?
That’s the long-term goal. But short term: if you’re a website selling GPU services for models, human developers buy your services today. Why aren’t you enabling agents to buy and upgrade the service as well?
There’s a time when the developer is offline. The service is at 90% and alarms are [00:09:00] going off. What happens next? Does the agent have the authority to buy some CPU or GPU power to keep itself up and running because the owner isn’t there?
John Koetsier:
Super interesting. Payments have continually accelerated with each platform shift. With this addition, it could be lightning fast — machine-speed transactions.
Jim Nguyen:
Absolutely. The goal is to remove the complexity of paying. Imagine if at a grocery store they said, “We can only take cash you withdrew five minutes ago.” That’s friction. But if they say, “Here’s your bill, we take any payment option,” that’s streamlined.
In the web era, payments were critical. In mobile, too. But the blockage in AI payments isn’t lack of payment capabilities. We have enough to get started. Our expertise is enabling frictionless money movement — not only at the UI level, but at the protocol level, whether fiat or stablecoins. [00:10:00] We focus on removing friction from moving and taking money.
John Koetsier:
You can’t get away from money, can you? You created Ruby Coins for in-game payments, then got bought by PayPal. Now you’re doing another thing around payments. You seem to love money.
Jim Nguyen:
I love the idea of commerce and productivity. When productivity happens, there’s a reward. Every platform shift brings new opportunities to build payments infrastructure.
My payments journey started before Ruby Coins. I was playing video games and running tournaments. My team members would be late. “Why are you late?” “Jim, the payment flow takes me out of the screen, I can’t come back into the game.” It happened to me too. I called my co-founder: “Can you do this?” He said impossible — then two weeks later showed me a demo.
That’s how Ruby Coins was born: [00:11:00] in-game payments where you never leave the game — just “top up” buttons.
We’ve always focused on delivering the best payer experience. Now we’re taking that understanding into agentic payments, where agents pay agents — with no UI, no forms.
John Koetsier:
So will an agent spend all my money? What are the guardrails?
Jim Nguyen:
Great question. Think about giving your child a debit card. You wouldn’t put all your money in there. You’d give them a hundred bucks. The statement comes back and you talk. Same with agents.
Humans have the logic to pay in their heads. How do you move that logic into a policy system to put guardrails around the agent? You can set: what sites they can go to, how much they can spend, time windows, etc.
Because it’s financial services, it must be regulated. Yes, there can be rogue agents, but in our system every agent is an extension of a person. [00:13:00] You sponsor the agent. You are liable. That linkage is a protective layer.
John Koetsier:
That’s built-in responsibility. Accountability. If something goes wrong, I know who to go to — there’s a paper trail.
It’s mind-blowing stuff. I talked to Savinay Barry, CTO of OpenText, about HR for agents — onboarding, end-of-life, retiring agents. Now financial systems will need budgets and policies for agent spending: how much, when, where.
It’s interesting — systems built for humans will need to adapt. One idea is an agentic marketplace where agents can hire, fire, pay, transact, farm out labor. Agents could advertise services, ratings, reviews, uptime, processing time, rate per unit of work. [00:16:00] You could onboard them easily, with APIs to communicate, send work, send money, create contracts.
Are you going to build that? Does that look like agent.ai by HubSpot? What’s it look like?
Jim Nguyen:
We focus primarily on monetization of each agent. Full stop. We work with marketplaces to onboard buyers and sellers in a regulated way, then enable someone like you or me to hire an agent to handle bills — with a budget and certain accounts they can touch.
As you gain trust, you give them more. We explore what’s possible but also believe marketplaces should operate the way they do today. We want to enable those marketplaces to connect agents: an agent buying a service from a human, or an agent selling to a human. Today it’s human to human — imagine adding a dimension where buyer and seller can both be agents. [00:17:00]
John Koetsier:
An AI agent can onboard a human, because that’s something an AI agent can do, or I need a real-world task done.
Gig marketplaces are going to change massively. A decade from now they’ll feel totally different. You’re saying you’re not going to boil the ocean — you’ll focus on the financial layer and work with marketplaces.
Jim Nguyen:
Absolutely.
John Koetsier:
Blow our minds. What does a fully autonomous work and commerce landscape look like in a few years?
Jim Nguyen:
Today, it’s about helping coding agents and workspaces alleviate friction. Next phase: enterprise integration. HR handles onboarding and offboarding. But agents could handle paychecks, vacation payouts, and other tasks below certain thresholds — with human approval above the threshold.
Andreessen wrote that almost every white-collar job will have an agentic platform. HR will have workspaces. Engineers already do. You may hire a QA AI agent to test your website constantly, or a company providing SEO via agents.
John Koetsier:
In this future world, agents are the new apps. You can publish and build agents — like GPTs on OpenAI’s platform. OpenAI doesn’t have payments yet, but I’m sure that’s coming.
Some aren’t happy about agents, though. Amazon told Perplexity’s Comet platform to keep its agents out. On Comet, you can build an AI agent and tell it to buy products at certain prices. Amazon said no — they have their own Rufus AI agent.
Is that a locus of control issue? Something else? How do you see this?
Jim Nguyen:
I’ve been on the edge of technology for a long time. Pushbacks often sound like “We don’t like AI agents,” but it’s not that. Marketplaces are optimized for humans — and optimized for revenue. Agents may shortcut those flows and shortcut the revenue model.
But once marketplaces learn how to optimize for agents, they’ll welcome them. Agents are a new customer — a buying customer. Companies need to level up and accept a new type of user.
Not so long ago, people asked why we needed stablecoins when we had US dollars. But stablecoins are now one of the primary ways to move money. Agents are the same — not a threat, just lacking the infrastructure.
John Koetsier:
Yep, that’s the missing piece.
Jim Nguyen:
Exactly.
John Koetsier:
In a recent Tech First, we talked about ghost kitchens — kitchens that exist only for Uber Eats. I think we might get headless e-commerce platforms: no website, no UX, no brand — just stuff, a warehouse, distribution, and agent-optimized APIs. Agents come in, talk, transact.
Jim Nguyen:
Exactly. That removes friction. Ordering takeout is a process today. But imagine saying to a voice agent, “Get me a burger with cheese, medium rare, by 7:30. Here’s my wallet.” And it’s done. Instead of going through all the steps yourself.
John Koetsier:
Jim, this has been super fascinating. Thank you for the conversation.
Jim Nguyen:
John, thank you for having me.