With AI systems becoming increasingly agentic (ie, capable of carrying out actions toward goals over extended periods of time, without humans being in the loop or pre-specifying their actions) and business-to-robot-to-consumer (B2R2C) commerce about to explode, it is a natural evolution for AIs to become economic actors in their own right and to exchange value in return for services. Bot-to-bot payments are building out a new frontier that will need new rails.
As McKinsey have said, chatbots are just the beginning of the evolution from the current knowledge-based, generative AI-based tools—such as chatbots that answer questions and generate content—to agents that use models to execute complex, multistep workflows across a digital world, “moving from talking to doing”. These software robots powered by agentic AIs — actually, for simplicity, let’s go back to calling them bots — will be just like human Internet users but much faster and much smarter.
As I have said, the big change in financial services will come when customers have bots, not when financial institutions have bots. It is not at all clear to me that banks, retailers, airlines and other have taken on board what this will mean for their medium-term strategies, and all of them will have to.
Take the example of airlines, one familiar to most readers. Every time I try to use my frequent flier miles to go somewhere interesting, there are no seats available. In order to get seats that are actually useful, I have to log in every day to the airline site and look a flights months from now, with the family calendar open on one side and my work calendar open on the other side) and check half a dozen different routes. Frankly, I have better things to do. But my bot doesn’t.
I’ve had more than one flight voucher expire because I couldn’t find a flight to use it, I’ve more than once abandoned a conference because I could use points to get there and I find myself increasingly frequently using other airlines because the points on my airline are declining in utility. How will the interaction between my travel bot and various airline bots change my behaviour, the airline’s profitability and the marketing strategies of its partners?
(I genuinely don’t know, which is why I am asking.)
I can see how a little downstream my bot could pay a British Airways bot using open banking and instant payments (and Stripe already have an SDK for AIs to pay using virtual card numbers), but it is hard to see how my bot could pay your bot a quarter using Visa or how my bot can pay some Substacks while it assembles a morning newsletter for me. This is why the bot-to-bot space has been so energised by the rise of stablecoins, giving bots the ability to pay each other a Pound or a Euro or a Dollar instantly for next-to-nothing means a fundamental change.
A post-industrial revolution needs post-industrial money. If the idea of bots paying other bots in the consumer space may seem a little distant, take a look at the enterprise space and consider how a global manufacturing firm could transform its supply chain operations using bots and post-industrial money. This goes beyond simple automation to produce cognitive integration that can learns and adapt to changing business conditions. It is very small step to give those agents the ability to not only initiate orders but also to pay for them!
Rather than simply linking inventory systems to procurement platforms, businesses could deploy orchestrated networks of intelligent agents that autonomously monitor stock levels, predict demand patterns, initiate orders and pay for services without human intervention or credit risks.
This is such an interesting new field for the payments industry to explore. I’ve co-written a paper on this with Debbie Gamble, chief strategy officer at Interac, in which we identity the potential for a new payments infrastructure to both re-energise past propositions (eg micropayments) and create entirely new ones (supply-chain currencies). New transactions and new trade mean new prosperity so with the right governance in place, the payments industry can explore this entirely new frontier to the great benefit of the economy as a whole.
The paper is called Agentic commerce and payments: Exploring the implications of robots paying robots and it has just been published in the Journal of Payment Strategy & Systems 19(1), p.72-84 (Spring 2025) and you’d be mad not to subscribe to this august journal of record and read it now.
There are startups already beginning to explore this new landscape. Nevermined (which raised $3m in its seed round) is building just such rails with the intention of becoming the “PayPay of AI Commerce”. Don Gossen, a co-founder of the company, told me that he believes that “AI agents will change commerce” and the fact is that he is right. We are seeing the birth of AI-driven economy that will be dominated by machine actors, with agents paying other agents directly for the services they require, and Nevermined sees an opportunity to provides the critical payment infrastructure needed to enable a new business paradigm.
Start-up Skyfire has created a payment network specifically for agents to make autonomous transactions. Their system assigns each AI agent a digital wallet with a unique identifier, where businesses can deposit a set amount of funds they want the agent to spend, so they don’t get unlimited access to a bank account. Skyfire also allows customers to set limits on how much an AI agent can spend in one transaction and over time. If an AI agent tries to overspend, it will ping a human to review it. Skyfire also offers a dashboard to view exactly how much, and where, their agent is spending.
Interestingly, when I asked Craig DeWitt of Skyfire how he planned to monetise a system of decentralised frictionless payments where the transactions margins will approximate to zero his answer was immediate. “Where we think we can make money is identity” he told me.
(A man after my own heart, frankly).
This is a crucial perspective. AI transactions are not only just about speed. They are not only about reliability and efficiency too. They are also about security: with the ability to integrate digital identity, agents ought to be much harder to fool than people are! The agents will ensure that all transactions comply with all the necessary security protocols while optimising the process to save time and resources. A person might not be able to tell whether their counterparty is really a Norwegian lawyer with a Manchester City season ticket, but their bot will be able to request credentials, resolve certificates and force authentication of private key whenever necessary.
I think this is where there might be opportunity for the incumbents. Banks might not be able to control the post-industrial rails or give me the post-industrial money to run on them, but they do know who I am and they do know who I pay and they do know how much my business owes them and so on. Jelena Hoffart, Director of Identity Value Chain Expansion at Mastercard, told me that she thinks it may turn out to be a pretty good business for the financial institutions to move beyond know-your-customer (KYC) to provide know-your-agent (KYA) identification and authorisation services and I think she is right.
In fact, I think that she is so right that we’re written a paper about this for the Journal of Digital Banking. It will be published later this year, so not only are agent-to-agent payments a thing, KYA is now officially a thing too.
Banking 4.0 – „how was the experience for you”
„To be honest I think that Sinaia, your conference, is much better then Davos.”
Many more interesting quotes in the video below:
