By Dana Malman Warren and David Hirsch
Summary
The consumer-agentic commerce story is exciting -- who doesn't love the idea of a commerce clone haggling on their behalf for the latest brand collab, and optimizing a custom closet upgrade that seamlessly refreshes their look without breaking the bank? With that said, the opportunity is hiding in plain sight, and read on to learn more about why we have conviction here.
An introduction and reminder:
In Dana's last two posts, she traced the arc from stablecoins as a new money movement super highway to the broader promise of intelligent commerce, where AI agents and programmable money converge to rewire how parties coordinate and pay to move value through our economy. If you've been following along, you know she believes this shift is real, structural, and already underway.
But there's a chapter in this story that doesn't get nearly enough airtime. It's not the consumer checkout experience. It's not the AI shopping agent buying you sneakers. It's the type of payments that quietly, expensively, and, to be frank, poorly run the global economy. We're talking about B2B payments.
Enter David, a former colleague of Dana’s from back at American Express, who has spent 20+ years in B2C and B2B payments. After cutting his teeth on payments at Amex, where he drove multi-billion dollar revenue initiatives in merchant pricing, David transitioned to Fintech, ultimately going on to Klarna to lead corporate development and later to Amazon, where he led their corporate and business development team focused on payments/FS and retail tech (B2B and B2C). There, he crafted multi-pronged partnerships with the likes of Affirm and Stripe. David served on the board of Tabby, the MENA BNPL, which has grown to a >$4Bn valuation as the region’s most valuable fintech. Now based in London, David is an active angel investor and advisor to fintechs in the region andhe and Dana are on a mission to prove that agentic commerce’s highest value-add potential sits squarely in B2B commerce.
We agree this is where the most durable, high-value agentic opportunity lives. And it's hiding in plain sight.
Why B2B Payments Are Perfect Agent Territory
Here's the paradox at the heart of B2B payments: the transactions are large, the stakes are high, and yet the underlying work is often highly routinized (apologies to anyone who adores this process. It's critical work! ). Compared to, say, shopping for something for yourself, though, the work of B2B commerce and payments is less centered on discovery and more on control, compliance, and visibility. And it’s governed by the management of a consistent, multistep process that ensures those 3 goals are met.
In fact, payment is generally fully decoupled from product discovery and selection, as that work has been done by the business and procurement teams long before the payables manager processes the invoice. Think about what actually happens when a mid-market company pays a supplier.
- Someone has to reconcile the invoice against the purchase order.
- Check the payment terms… net 30, net 60, early-pay discount windows.
- Validate the counterparty.
- Route through the right approval chain.
- Choose the optimal payment rail.
- Manage FX if there's a cross-border leg.
- Log it.
- Reconcile it.
Then, repeat hundreds or thousands of times a month.
This is not work that requires creativity or judgment at every step. It requires precision, consistency, and the ability to keep many variables in mind simultaneously. These are things agents do extraordinarily well. The same qualities that make B2B payments tedious for humans make them well-suited for automation.
We saw this pattern quite clearly in earlier B2B payment innovations, such as vPayment at American Express, which was the first tool to leverage virtual card numbers at scale to ensure compliance from PO through to payment. Similarly, at Amazon, we witnessed growth in E2E procure-to-pay SaaS as smaller businesses sought to use virtual cards for their purchases to facilitate smooth reconciliation. While these solutions helped automate individual steps and establish/ensure workflows, agentic solutions allow for full process automation, removing the endless number of human clicks to tick boxes, paste VCNs, and upload documents. That means meaningful operational savings from reducing skilled labor for rote tasks.
But the true value creation potential in agentic B2B payments goes beyond simple automation to a dynamic modeling layer. Optimizing payment timing across a supplier portfolio by balancing cash flow, capturing early payment discounts, and managing DPO. This is all a complex, multi-variable, high-impact problem. Getting it right can move meaningful dollars for a company. Getting it wrong creates working-capital drag, damages supplier relationships, or misses discount windows. Rather than riding an asymptotic approach to cost zero, the potential value is unbounded. An agent that can model these trade-offs in real time, learning from each cycle, is not just an efficient play; it has the potential to unlock massive working capital, inform supplier optimizations, and transform cash flow management.
We truly believe this agent or platform will become the financial edge.
Agents Aren’t Enough: the Network Problem Is the Real One
We know why we are not seeing all too many builders pick up pace and scale in this space: the technology is the easier part.
The hardest and most important challenge in B2B agentic payments is not building a smart agent. It is building a network. Payments are bilateral by definition: every transaction involves a buyer and a supplier, but in the case of B2B payments, that base of suppliers is highly fragmented, each with different processes for buyers to follow, and often they do not accept standard payment cards. An agent that optimizes on behalf of a buyer is only as useful as its ability to reach and interact with the supplier on the other end. If the supplier isn't receptive to the agent (i.e., “on the network”), the agent hits a wall and falls back to manual processes, undercutting the potential for value creation.
This is the longest pole in the tent. The companies that will win in agentic B2B payments are not necessarily the ones with the most sophisticated AI tools. They are the ones that figure out network build-out first, fastest, and most scalably. They get buyers and suppliers onto the same rails, use that interface to create genuine two-sided value, and, in doing so, create the kind of switching costs that compound over time.
Of course, we love and appreciate a two-sided network problem: we grew up professionally at American Express!
In fact, we've seen this movie before. It took Visa decades to stitch together merchants, banks, and consumers before the network effects became self-sustaining. The stablecoin layer Dana wrote about earlier creates a meaningful opportunity to build new, more efficient rails that underpin new network(s), but these still need to be assembled, relationship by relationship, vertical by vertical.
And the particulars of the vertical matters here. Payment dynamics in construction (milestone-based, heavily disputed, slow to settle) are entirely different from those in healthcare (insurance adjudication, remittance complexity) and from those in professional services (retainer structures, project-based billing). Solutions that build into a specific vertical (and founders who take the time to deeply understand stakeholder needs in that vertical) before going horizontal will have a structural advantage in network density, product fit, and the flywheel of network effects that drives scaled adoption.
The CFO Stack Is Ready… Promise!
Something meaningful has shifted in the last 18 months. CFOs, long among the most skeptical adopters of new technology, are now building genuine comfort with AI-powered workflows. And their BODs demand it!
The CFO stack today is a patchwork of point solutions and comprehensive platforms: ERP systems, treasury management tools, procurement software, accounts payable automation, spend management, and FP&A layers. The list is long, and the integrations are often fragile.
What's changed is that AI is no longer seen as a threat to the finance function; instead, it's being pulled in as a force multiplier. Anthropic's financial services focus, along with the rapid deployment of AI copilots across the enterprise software landscape, has normalized the idea that agents can handle complex, rules-based financial work. CFOs are asking the right question now: not "should we trust AI with our money?" but "which workflows are best suited for agents to run?"
The answer, more often than not, points to B2B payments.
Where This Is Heading, and why we’re so excited
We are still early, but frankly, we wish it weren’t so early!
Most of the agentic B2B payments activity today sits in narrow automation like invoice processing, payment scheduling, and approval workflows. The higher value potential work, agents that model payment strategy across a supplier portfolio, negotiate terms dynamically, and execute across programmable rails with embedded compliance, is just beginning to emerge.
What gives us conviction that now is the time is the simultaneous convergence of means, supply, and demand: programmable money infrastructure that makes new rail architectures feasible (“means”, a generation of founders who understand both the payments complexity and the agent design challenges required to solve it (supply), and CFO organizations that are now genuinely open to AI-led workflows (demand).
The consumer agentic commerce story is exciting and relatable. But B2B payments are where the dollars are largest, the problems are most persistent, and the opportunity to build a truly durable agent-powered network business is most compelling. It is, by nature, unglamorous work. And that's exactly why we're paying attention.
Dana is actively looking for founders building in this space, and David is keen to help those founders get traction on product and go-to-market. If that's you, run, don’t walk, to reach out!