Why CFOs Should Look to Healthcare on Agentic B2B
Payments companies already know how to authenticate machines. API keys, tokens, certificates and digital credentials can establish that a request came from an authorized system. But identifying a machine like an artificial intelligence agent does not answer the more economically important question: What is it allowed to decide?
And across enterprise ecosystems where the context and data workflows surrounding a payment can be more important than the payment itself, that question of permission is becoming, if not all-important, nearly all-important for B2B CFOs and finance teams.
But this week’s headlines show how the healthcare space is turning into a surprisingly useful laboratory for the future of B2B payments.
Forus, an AI network for medicine, announced a $150 million in Series C financing at a $3 billion valuation on Sept. 8; while Verily Health Inc. announced a new investment from Nvidia alongside additional investments from its existing investors on Sept. 9 to help healthcare organizations make their data AI-ready and deploy AI responsibly across research and care.
After all, AI agents are beginning to perform administrative healthcare work that requires them to navigate precisely the problems enterprise commerce will eventually face across workflow soft spots like identity, delegated authority, policy interpretation, documentation, counterparty communication, exceptions and escalation.
Software isn’t simply being asked to complete a task. It is being asked to act on behalf of someone else inside a system of rules and leave enough evidence behind for another party to trust the result. That may be the missing infrastructure layer in agentic payments.
Read more: AI Agents Need Permissioned Funding Sources. Not Company Bank Accounts.
The Hard Problem Isn’t Agent Identity; It’s Agent Authority
AI agents are not simply another form of automation. Traditional software generally executes predetermined instructions. An agent may be asked to interpret an objective, choose among possible actions and adapt when circumstances change.
The week’s funding news shows how healthcare AI is beginning to operate inside workflows where software must interpret rules, gather evidence, communicate with counterparties, resolve exceptions and sometimes determine what should happen next. A binary permission such as agent permitted or agent denied is almost useless in that environment. Authority has to become contextual.
Consider a corporate procurement agent told to replenish inventory without interrupting production. Authenticating that agent does little to determine whether it can select a new supplier, accept a price increase, modify quantities, negotiate payment terms, exceed a department budget or initiate payment. Each decision could require a different level of delegated authority.
As a result, the competitive center of B2B software is beginning to shift toward permission architecture. Enterprises already encode authority throughout their systems. ERP platforms contain budgets and cost centers, procurement applications maintain approved-vendor rules, corporate cards enforce spending limits, contracts establish commercial boundaries and identity systems define organizational roles. Those controls look less like administrative bureaucracy and more like today’s raw material for machine-readable governance.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
The PYMNTS Intelligence report “How Agentic AI Went From Zero to CFO Test Runs in 90 Days” found in September 2025 that nearly 7% of enterprise CFOs in the United States had deployed agentic AI in live finance workflows, while an additional 5% were running pilots.
Read more: FTC Puts Digital Health’s Growth Playbook on Trial
The Transaction Is Becoming a Chain of Authority
Healthcare offers another useful lesson because the difficult part of automation is often not processing the normal case. It is governing the exception. An agent handling a straightforward workflow can follow established rules. But enterprise value often depends on what happens when those rules collide with reality.
Human organizations handle these situations through judgment and escalation. Agentic systems will need a machine-readable equivalent.
That suggests the next generation of payment controls may move beyond static limits such as “this card can spend $10,000.” A company might instead specify that an agent can autonomously transact with approved suppliers up to a certain amount, tolerate price variance within a defined range and modify shipping terms when the expected cost of delay exceeds the premium.
After all, just slapping an AI agent into a transaction is unlikely to result in a successful experience. According to the PYMNTS Intelligence report “The Power of Now: Moving at the Speed of Life: Why Real-Time Payments Matter for Healthcare Insurance Payouts,” a collaboration with Visa Direct, payment delays are shaping how consumers experience insurers after a claim has already been resolved.
There is also a final lesson healthcare can offer finance: autonomous systems need to leave evidence behind. Healthcare already operates in a world where administrative decisions can require documentation connecting information, policies, communications and outcomes.
Agentic payments may have to adopt the same logic.
“There are existing bodies of law that, while not passed or promulgated for the reason of AI, are still applicable to AI solutions,” Alaap Shah, member of the firm at Epstein Becker Green, told Competition Policy International, a PYMNTS company, in an interview this month, adding that the question for healthcare firms is not simply whether an AI-specific law applies but whether the system creates risks covered by older legal obligations.
For all PYMNTS artificial intelligence, B2B and digital transformation coverage, subscribe to the daily AI, B2B and Digital Transformation Newsletters.
