Every market has its semi-permanent operating assumptions. For B2B payments, complexity is what has been treated almost as a permanent feature of the market. There are just, quite simply, too many rails, too many exceptions, too many invoices, and overall, too much manual work.
“It’s physically messy,” said Billtrust CEO Grant Halloran in conversation with PYMNTS for the August 2026 edition of the What’s Next in Payments series, “Only the Paranoid Thrive?”
Buyers pay through different channels, suppliers operate across fragmented systems, checks persist alongside cards and digital rails, and the process of turning an invoice into usable cash remains stubbornly labor-intensive. That’s just the corporate payments’ modus operandi.
“Nearly 60% of invoices are overdue today,” Halloran said, estimating that at any moment there is between $1.5 trillion and $2 trillion of what he calls “trapped cash” sitting in the B2B economy.
With borrowing costs remaining elevated and days sales outstanding rising, that trapped cash is becoming harder to ignore. The conventional response has been to automate pieces of the order-to-cash process.
Halloran pointed to the new capabilities firms today have for deciding, continuously, how each dollar of receivables should be converted into cash without damaging the customer relationship that created it.
“Businesses are trying to generate the most cash possible from their receivables at the fastest rate and at the best economics,” he said. “Historically, the magnitude of what we’re trying to solve would take a long, long, long, long time, and that’s why it hasn’t been solved before. But in this new era with artificial intelligence … we’re able to actually dream big and then execute quite rapidly.”
And that makes all the difference in an environment where extracting cash from receivables is becoming less an administrative exercise than a working-capital priority.
AI Is Turning B2B Payments’ Data Messiness Problem Into a Strategic Opportunity
Instead of treating a portfolio of overdue invoices as a queue, companies can help firms distinguish among buyers by behavior, economics, relationship value and likelihood of payment. That changes collections from a standardized sequence of reminders into a decision system.
Halloran calls the payment itself the “magic moment” inside a much broader cash-generation process full of signals that can be used to make more intelligent decisions about how receivables should be managed.
“You can’t just force buyers into the cheapest rail possible,” he said, stressing that the receivables optimization problem is multidimensional: generate cash faster, improve economics and preserve the customer relationship.
After all, if collections were a technology problem, it wouldn’t exist. The mechanics of moving money are becoming easier. The harder problems in receivables sit upstream: companies often lack a precise, real-time understanding of why cash they have already earned has not arrived.
“The amount of data that we’re dealing with is quite profound,” Halloran said. “When you take a trillion dollars of trade credit commerce going through the platform, there’s billions of events, behavioral signals and nonbehavioral signals as well, that now bring a data science lens to all of this.”
AI Is Raising the Value of Human Judgment
For finance executives, that suggests the operating model may eventually invert.
Machines can handle more of the repetitive decisions around invoices, collections and payment routing. Humans become responsible for the exceptions where customer relationships, commercial context or strategic judgment outweigh the algorithmically obvious answer.
“The biggest inflection is going to be how do we enable the humans to be doing the most value-added work, in more strategic decisions, that directs the operational,” Halloran said.
A supplier deciding how to collect an overdue invoice is simultaneously making decisions about working capital, payment acceptance, customer experience and future sales. An algorithm may identify the economically optimal payment path. The relationship manager may know that insisting on it could jeopardize a valuable account.
The value of AI, in that construct, is not eliminating judgment. It is concentrating human judgment where the economics justify it.
Watch the full PYMNTS TV episode with Billtrust CEO Grant Halloran to hear more about:
- Why AI is turning collections into a revenue-management problem. Halloran argues suppliers cannot optimize solely for the cheapest payment rail or fastest collection because forcing buyers into unfavorable experiences can damage future sales. The emerging opportunity is to optimize cash generation and customer economics simultaneously.
- Why B2B receivables are becoming a data science problem. With billions of behavioral signals across credit, invoicing, collections and payments, Halloran sees AI enabling suppliers to make increasingly granular decisions about which buyers require automation, intervention or different payment experiences.
- Why the endgame is selective autonomy, not removing humans. Halloran says AI should push employees toward higher-value judgment while automating operational work, leaving humans to intervene when relationships, intuition or commercial context outweigh what an algorithm recommends.
