Sunlight CEO Guy Ziv told PYMNTS earlier this month that B2B card acceptance is higher than the industry assumes. The real problem, he argued, is not that suppliers won’t take cards — it’s that navigating which invoices can be paid by card, through which portal, on what terms, and at what cost is so fragmented that most buyers never get there.
He is right. And I can put a number on just how right he is.
At Finexio, we made a specific bet: handle every payment channel. Phone payments. Portal payments. Every single way a supplier would accept a virtual card. Every competitor we encountered routed around anything that required human intervention — they built for the easy path and left the rest on the table. We never found one that had built scaled capability for phone and portal payments.
Based on years of routing actual payment volume across our supplier network, we knew those hard-to-reach channels represented approximately 20% of all virtual card spend — a fifth of the market that almost nobody was serving. I am talking about Finexio’s observed share here, but the pattern held across every data set and partner conversation we had.
The math is straightforward. If you are running a virtual card program and you are not capturing phone and portal payments, you are structurally capped at about 80% of the available spend. You can optimize the easy channels all day — the ERP-integrated, straight-through flows — but you are leaving roughly a fifth of your potential volume behind.
In every bank conversation I had at Finexio, I never found one that offered phone and portal payment services as a real, scaled capability. We talked to dozens of them. None of them had built it. The same was true for most payment providers. The industry has treated phone and portal payments as a niche problem, the long tail of B2B payment acceptance, not worth automating because it is too hard to automate at scale.
That was a defensible take five years ago. It is not defensible today.
Because here is where the Sunlight framing connects directly to the operating reality: computer use via AI is going to be a major unlock for exactly this problem. When an AI agent can navigate a supplier portal the same way a human AP clerk would — enter the invoice number, check the box, enter the card details — the cost of serving that phone-and-portal channel collapses. It becomes automatable at scale.
This is not a futuristic thesis. The underlying technology is already live. Agentic AI systems can see, click, fill, and submit in browser environments. The only thing that has been missing is the specific training data and workflow logic for thousands of supplier portals — and that data exists, embedded in the historical payment records of every company that has been doing this work.
The opportunity gets overlooked because the industry debate on virtual cards has been dominated by a different question: “Will suppliers accept cards?” Sunlight is doing a service by reframing this. The acceptance data is better than anyone thinks. The real bottleneck is execution — can you actually complete the payment across every channel your suppliers use, given every portal variation, every unique workflow, every different set of rules?
When you look at it that way, the trapped volume starts to look different. It is not 20% that is inherently unreachable. It is 20% that is currently trapped in manual processes that AI is about to automate. Over time, that is an embedded ~20% yield improvement in virtual card programs — recapturing a fifth of the spend that is already in your network but not yet captured at the payment rail.
For an AP platform or a payfac, the call to action is simple: audit your supplier network data now. If you are only routing card payments through the easy channels, there is rebate yield sitting in your data that you are not collecting. Your own payment records show which suppliers already process cards through portals or phone lines — you just have not built the infrastructure to reach them.
The providers who figure out the phone-and-portal problem with AI agents will capture that spread. Everyone else will keep fighting over the 80% that every competitor can already reach.
This is what I mean when I say financial services is pipes and plumbing. The sexy problems get the press. The boring operational problems — which portals take which cards, on what terms, with what data — are where the real value is buried. And AI is about to make that plumbing dramatically cheaper to build.