TBPN · 8 min · Jan 27, 2026

Agent payments need better economics and a clear use case

AFAaron Frank · Partner, Lightspeed

Aaron Frank argues that agentic commerce needs to lower costs or improve conversion, and merchants still lack ways to trust agent identities.

The episode’s best minutes

3:05:17 A personal assistant emerged from a simple messaging hack6 min
1:37:41 Media companies lose their way when they chase scalePro pick 7 min
3:38:31Agent payments need better economics and a clear use casePro pick 8 min
5 more moments in the full episode

Up next · SourcesChatGPT is starting with intent-based ads and firm limits8:00

“Agentic payments have to actually be more efficient for the system to be adopted. Otherwise, we're just going back to card.”

AFAaron Frankat 3:39:48 · TBPN

Takeaways Pro pick

Agentic payments must beat card rails on efficiency to gain adoption.
Merchants need ways to verify agents and assess fraud risk.
Stablecoins can give people worldwide access to US dollars.

About this segment

Aaron Frank argues that agentic commerce needs to lower costs or improve conversion, and merchants still lack ways to trust agent identities. He sees stablecoins’ clearest use as making US dollars accessible globally.

AF

Partner, Lightspeed

John Coogan, Host · Co-host, TBPN
Co-founded Soylent and Lucy; OpenAI bought his show TBPN in April 2026.
Jordi Hays, Host · Co-host, TBPN
Founded Branded Native; co-founded fintech Capital, which Rho bought.

From “🦞 Moltbot / Clawdbot Creator Peter Steinberger, Palmer Luckey, and more”, Jan 27, 2026. Watch the full episode

Transcript

From the transcript TBPN publishes with the episode.

3:38:38
It's a agent to commerce is a weird thing. Right? Like, actually don't have the payment rails to enable agents to proliferate. Oh, man. Proliferate. Yep. We've funded some people doing that, building kinda l ones to essentially provide that next rail. But the reality is the biggest distribution networks are Visa, Mastercard on the merchant side. And they don't have really a cost structure that sorry. They have a cost structure that can support it.
3:39:01
They just don't have a market force to that forces them there. Stripe and Tempo launching well, Tempo is not live yet, but supporting kind of a new network there. Ultimately, stablecoins may be, you know, the path there for at least how we transfer value, but it's also just kind of a store of value right now. Agent to commerce is just an inning one because we also just don't have a use case. The joke I make is I haven't I haven't actually been able to buy anything with an agent in the news.
3:39:27
This week was what Shopify is tacking on 4% on top of transaction, which They're sending it to OpenAI, but Yeah. Yeah. Yeah. Yeah. But like that just that's untenable for a merchant. Right? Some merchants. I mean, there's plenty For dropshippers, it's fine, but like but for any, like
3:39:39
There are plenty of merchants that have 20% coupon codes in every podcast. That's true. But so but that's a very small portion. If you think about like Walmart, Walmart literally doesn't take Apple Pay because it doesn't wanna pay the 15 extra or whatever it is points. Very good point. So like when you get to large trillions of dollars of commerce Yeah. Agentic payments have to actually be more efficient for the system to be adopted. Otherwise, we're just going back to card.
3:39:59
Because you need a lift in actual consumer behavior. Consumption needs to increase to offset that 4%. Correct.
3:40:05
E commerce needs either basket size increase or checkout conversion increase in like
3:40:09
I was I was worried about a situation where somebody discovers a product on Instagram Mhmm. Via an ad that a brand paid for, and then they just go in a chat GBT Say order this for or another model and say like, order this for me, and then it's like and just an incremental 4%, you already paid to acquire the customer effectively. Yeah. And depending on how great this integration is, it might be that consumers should start to prefer that.
3:40:33
Yeah. Yeah. It it's an evolving field. There's a whole another thing is how do you trust an agent if I'm Lululemon and they just show up? Like, I don't know whose agent this is. I don't know if it if it's personal or it's a business. Mhmm. And so, we don't actually have a layer in the Internet today that says, you can trust this agent, it's not fraud. Which if you think about how an e commerce checkout store works today Mhmm.
3:40:54
They have multiple, you know, consumer puts in a card, there's multiple layers of fraud checks that happen. And so when an agent shows up from the same IP for 50 times in the same day, you have no idea what that consumer is. And then you lose all the downstream tracking that you do on a consumer Yeah. Per transaction. Yeah. I mean So it just it breaks the entire model of the Internet.
3:41:12
Yeah. Was I was I was sort of cautiously optimistic that that agent to commerce would roll out before Black Friday Yeah. Towards the at the the end of last year. That didn't really happen. I mean, I've had the same experience as you. I don't think I've purchased anything agentically.
3:41:28
I've tried to buy flights and even search is bad for it. So we just need to build more and more protocols.
3:41:32
I think the I I mean, I think the the the current, like, use case or or, like, the would just be, you're trying to buy a single product. Mhmm. And not a complex cart yet, just something simple. And it I think it can get across the finish line. But, again, I haven't actually been, like, moved to do that. And I'm wondering where GMV on agentic platforms will land at the end of this year. It feels like there's gonna be a massive push. All the integrations are happening. All the deals are happening. Like, you can see the UI is is adapting to it.
3:42:03
Yeah. I think OpenAI already has my credit card in ChatGPT. They should be able to do this. But I'm wondering, like, consumer behavior, awareness, education, there's a lot in there. There's an images tab. There's a deep research tab. Like, there's just a lot that could put, like, a like a slowdown on the roll out there.
3:42:18
Yeah. There's an old saying in payments is that everything happens over a ten year time horizon. AI clearly is shifting that shorter, but Yeah. Like, until it's actually saving you time and you trust that they're doing they're getting you a discount, they're getting you the most efficient price or something like that Yeah. It dot consumer behavior is fundamentally the hardest thing to change. Totally. So there's there's it it will exist.
3:42:38
Yeah. I'm sure there are people buying stuff. I'm sure, like, I will get slammed for saying, like, who's done an agentic transaction? But the reality is it's I mean, are out there giving their
3:42:46
The classic thing last year was, like, I just just let me order a flight, you know. And I think we're now we're now getting We're close.
3:42:57
And we've we've had agent ecommerce. There there was when we were running final, we had a ton of people using the product for sneaker bots. Oh, okay. And and so it's it's a bot, you know, agentic commerce is bots by another name.
3:43:07
Yeah. I guess that is agentic commerce technically. But Yeah. It's just like you now have this layer of intelligence that an LLM gives you. Yeah. So it can do complex coordination and complex tasks.
3:43:16
I had an auction sniper in like 2004 on eBay. Yeah. Click the button right before the auction ends.
3:43:22
They still exist. They still exist. And there's a ton of volume flow, but you know what? It still still runs over the card rails because it's the lowest common denominator.
3:43:29
Yep. Yep. That makes sense. Interesting. So is is one of the strongest bull cases for stablecoins still just lower transaction fees and No. Not at all.
3:43:37
No? No. There's no domestic man, I'm gonna get slammed for this. Oh, yeah. Yeah. So so it doesn't seem like it doesn't seem like any any of the the players that have the ability to get institutional adoption of stable coins Mhmm. Have any incentive to be like, oh yeah, like, it's gonna be like point o 1% fees. No. Why would you why would you you maybe wanna be competitive Mhmm. But it's there's no incentive to make it zero or close to zero.
3:44:02
You would do zero so you get the other side of the equation. If you think about like a a store of value, like if you look at like a dollar Yeah. I give you a dollar, you know it's worth a dollar. There's no risk transfer there at all. Yeah. Yeah. And that's actually what what we want in in payments. And it's actually why, know, Visa, MasterCard are the largest factoring networks in the world. When you swipe at Starbucks, they don't get a dollar, they get 97 and a half cents on the dollar because that is the risk that they were taking in taking your dollar. Yeah.
3:44:30
Stablecoins, you would give it away for free. One, because ultimately, Visa Mask for man, we're going way deep in payments, but Sure. Are actually data networks. Yeah. They actually don't transfer money. The money is transferred via settlement layer at Fedwire. Yeah. And when you think about it, like, how would you do with stable coins? Well, it's still just a data layer and you actually just wanna hold on to as much treasuries as possible Got it.
3:44:51
And make the yield from that. And so how would you aggregate volume? Will you give away the transfer for free? Because data, in theory, as if you look at a bunch of network effect things, as data goes up to the right, the cost of that data should come down to the right. And so you would just want as much volume over your network to win. And that's why you have things like Tempo launching. You have the other kind of l ones launching to try to aggregate all that data, while you also have the incumbents who have a lot of money and a lot of market force to be able to try out new things to try to capture that volume back.
3:45:17
Were you able to stop launching new l ones? There'll be new use cases. You can launch them. Whether you get adoption is a different question. Sure. Right? Like, we backed one called Radius Technology, building really interesting technology as an l one. It enables transaction volumes, I think Bobby would say, in the millions. Yeah. They came out of the Fed, kind of the the central bank digital currency project that died. They turned project Hamilton into this company.
3:45:42
Okay. And in doing so, yeah, if you wanna do a giant e commerce, you're gonna be talking about millions of transactions a second. Yeah. And you can't do that over existing rails Sure. In a way that doesn't also break a bunch of other primitives. So, yeah, we'll keep on launching l ones. What I was saying about stablecoins is there's really stablecoin is just kind of a quasi bank account. Right? We finally have true programmatic money.
3:46:03
Behind the scenes, there's a lot that happens to on ramp and off ramp. But what it enables us to truly take what is The US's biggest export, is the US dollar by far Mhmm. And export it globally without any barriers. Mhmm. So if I'm any other country, friendly or not, my citizens can now get access to the US dollar. Yeah. And that is that is the demonstrable use case of stable coins. Everything is else downstream of that.
3:46:26
Mhmm. Yeah. Like Venezuela is a great example right now. Do you trust your central currency? Who is your like, no. You don't even trust the central bank, so I'd rather hold on to cat, US dollars, which