PNC’s FedNow Move: What It Means for AI-Secured RTP

A single operational decision by the U.S. Treasury just did what two years of “real-time payments are the future” messaging couldn’t: it pushed a major bank off the sidelines.

In October, PNC joined FedNow—more than two years after the Federal Reserve launched the instant payments rail. The trigger wasn’t a flashy product feature. It was a use case with consequences: Treasury signaled disaster payments would be delivered over FedNow. PNC’s logic was blunt and practical: if federal relief can arrive instantly, a retail bank can’t be the one explaining why its customers have to wait.

For anyone building payments infrastructure—banks, processors, fintechs, and payment ops teams—PNC’s move is a case study in how modernization actually happens. Real-time rails aren’t adopted because they’re exciting. They’re adopted because the ecosystem (government, employers, merchants, and customers) makes waiting unacceptable. And once money moves in seconds, AI becomes less of a nice-to-have and more of the control layer that keeps the system safe, routed correctly, and operationally sane.

Why PNC joined FedNow (and why it wasn’t “FOMO”)

PNC joined FedNow for two reasons: mandatory relevance (disaster payments) and network reach (getting closer to 100% coverage).

At a Federal Reserve virtual town hall in December, PNC’s treasury management payments leader Sarah Billings described the inflection point: when Treasury signaled it intended to pay disaster aid to individuals using FedNow. The implication is obvious if you run retail payments: customers don’t compare your payout speed to your branch down the street anymore—they compare it to the moment they’re told help is coming.

Disaster payouts changed the ROI math

Disaster payments are a rare payments product requirement that hits all the pressure points at once:

• Speed matters emotionally and financially (rent, food, temporary housing).
• Volume can spike unpredictably (storms, wildfires, floods).
• Fraud attempts surge (impersonation, account takeover, synthetic IDs).
• Public scrutiny is high (the “why can’t I get my money?” story spreads fast).

That combination turns “real-time payments” from a roadmap item into a risk and reputation issue. PNC’s quote captured it: as a large retail bank, it couldn’t imagine not giving clients access to immediate funds when they needed it most.

Coverage matters more than rail preferences

The second driver is less emotional but just as important: reach.

Before FedNow, PNC said it could reach about 80% of banks with real-time services through its existing options (including The Clearing House RTP network). But “80% reachable” still means “1 out of 5 counterparties can’t receive instantly,” which is a terrible experience for payroll, insurance payouts, gig work, and consumer transfers. FedNow offered a path toward broader connectivity—assuming adoption continues.

This is the infrastructure reality most executives underestimate: the best rail is the one your counterparty can actually receive.

FedNow adoption is growing—but the long tail is the real battleground

FedNow has signed up roughly 1,500 banks and credit unions since launching in July 2023, out of about 9,000 U.S. financial institutions.

That number tells two stories at the same time:

1. Momentum is real. You don’t get to 1,500 participants on pure hype.
2. The hard part is ahead. The remaining institutions include many smaller FIs with lean tech teams, limited fraud staffing, and vendor constraints.

A Fed official noted that current FedNow users range from about $500 million to $3 trillion in assets—meaning the network has both small and very large participants already. But scale isn’t just “more banks.” It’s more endpoints, more exceptions, more account types, more fraud patterns, and more operational edge cases.

Here’s the stance I’ll take: real-time payments don’t become mainstream until onboarding the long tail is boring. That requires better tooling, better monitoring, and better automated decisioning—aka, AI applied to payments operations.

What changes when payments are instant: risk, ops, and customer expectations

Instant payments compress time. That sounds like a product upgrade. It’s also a governance upgrade.

With ACH, you can often detect anomalies, reconcile, and respond on a human timeline. With real-time payments, you’re making decisions in milliseconds—and once the money is gone, you’re in recovery mode.

The “no-back-button” problem

Real-time rails are close to irrevocable in practice. That changes the entire risk posture:

• Pre-transaction fraud detection matters more than post-transaction investigation.
• Identity confidence must be evaluated continuously, not just at login.
• Operational controls need to work 24/7/365, including holidays.

PNC highlighted a very human business scenario: a corporate client forgets to process payroll (especially around bank holidays not everyone observes). Real-time payments can fix the problem instantly—but only if the bank is comfortable sending those funds instantly.

That comfort comes from strong controls and fast decisions. This is where AI belongs.

Where AI fits in FedNow and real-time payments infrastructure

AI’s value in real-time payments is straightforward: it helps you make correct decisions fast, at scale, under uncertainty.

Not “AI for AI’s sake.” Practical AI that reduces fraud losses, reduces false positives, and improves routing and availability.

AI for real-time fraud detection (the non-negotiable use case)

Real-time fraud detection has three jobs: score risk quickly, explain decisions, and adapt to new attack patterns.

Effective models typically combine:

• Behavioral signals (device changes, typing cadence, session anomalies)
• Payment graph signals (new recipients, unusual velocity, hub-and-spoke patterns)
• Customer context (normal payees, historical amounts, time-of-day norms)
• External intelligence (known mule accounts, compromised credentials)

The key is not just building a model—it’s deploying it in a way that works with instant rails:

• You need sub-second scoring.
• You need human-review fallbacks that don’t block legitimate urgent payments.
• You need policy controls that degrade gracefully when systems are under load.

A memorable rule for real-time rails: false positives are customer churn; false negatives are headlines. AI helps balance that tradeoff when volume spikes—like during disaster disbursements.

AI for payment routing and reach optimization

PNC’s “reach more institutions” point is the routing story in disguise.

As banks connect to multiple rails (FedNow, RTP, ACH same-day, wires), routing becomes a decision engine:

• Is the recipient reachable on Rail A right now?
• What’s the cost and SLA on Rail B?
• What are the customer’s preferences and limits?
• Is fraud risk elevated for this rail/recipient pair?

AI can assist by predicting successful completion probability and recommending the best rail given constraints. Even a simple machine-learning model that forecasts failure rates by endpoint, time, and message type can reduce retries and customer support tickets.

AI for 24/7 operations (because humans sleep)

Instant payments are always on. Many bank operations teams aren’t.

AI-driven ops isn’t glamorous, but it pays off quickly:

• Automated anomaly detection on settlement positions
• Predictive alerting for queue backlogs or interface degradation
• Intelligent case triage (grouping related payment incidents)
• Automated customer communications for known incidents

This matters in December more than most months. Year-end brings higher volumes, more payroll irregularities, more account changes, and more staffing constraints. Real-time rails don’t care that it’s holiday season.

What PNC’s decision signals for banks and fintechs in 2026

PNC joining FedNow is a signal that use-case gravity is stronger than rail debates.

It’s easy to argue FedNow versus RTP on paper. It’s harder to tell a customer they can’t receive federal disaster funds instantly because your institution didn’t prioritize a connection. Government disbursements (and large enterprise payroll and insurance use cases) create a forcing function.

Expect “real-time readiness” to become a competitive baseline

When enough institutions can receive instantly, customer expectations reset. Then every laggard looks broken.

For banks, “real-time readiness” will increasingly mean:

• Access to at least one instant rail (and often more than one)
• A routing strategy that prioritizes completion and customer intent
• Strong instant-fraud controls with measurable performance
• Clear customer experiences for limits, holds, and exceptions

For fintechs, it means your platform has to support:

• Multi-rail connectivity
• Instant reconciliation and status visibility
• Fraud tools that operate at real-time speed
• Audit-ready decision logs (especially for AI-assisted decisions)

The winning teams treat AI as infrastructure, not a feature

Here’s what works in practice: don’t bolt AI onto the side of payments. Put it into the core loops—risk scoring, routing, monitoring, and exception handling.

If you’re implementing FedNow (or expanding instant payment capabilities), build an “AI control plane” roadmap alongside the rail connection roadmap. Otherwise you’ll ship speed first and spend the next year fighting fraud, false declines, and operational burnout.

A practical checklist: deploying AI for FedNow safely

If you’re responsible for payments infrastructure—at a bank, processor, or fintech—this is the checklist I’d want on my desk before scaling instant payments.

1. Define your real-time risk appetite in numbers

   • Target fraud loss rate (bps)
   • Target false positive rate
   • Target manual review rate

2. Instrument every step of the payment lifecycle

   • Request received, validation, scoring, decision, send, confirm, post
   • Latency tracking per component (model, rules, rail gateway)

3. Use layered decisioning (rules + ML + customer controls)

   • Rules for known hard blocks
   • ML for pattern detection
   • Customer-configurable controls for limits and recipients

4. Build explainability into the workflow

   • Store top drivers for each model decision
   • Make it readable for ops and audit teams

5. Plan for surge events (like disasters)

   • Predefined “surge mode” thresholds
   • Additional step-up verification paths
   • Automated comms templates and status pages

6. Close the loop with outcomes

   • Label fraud, disputes, returns, and customer complaints
   • Retrain and recalibrate models on real outcomes

Fast rails punish sloppy feedback loops. Tight loops win.

The question to ask after PNC’s FedNow decision

PNC’s FedNow adoption wasn’t about novelty—it was about being ready when instant money becomes the default, not the exception.

In the broader AI in Payments & Fintech Infrastructure series, I keep coming back to the same point: as payment rails speed up, decisioning has to speed up too. AI is the most practical way to do that without either (a) drowning in fraud, or (b) blocking so many payments that customers stop trusting the product.

If you’re planning your 2026 roadmap, the forward-looking question isn’t “Should we connect to real-time rails?” It’s this:

When instant payments spike—because of payroll mistakes, market volatility, or disaster relief—do you have an AI-driven control layer that can keep approvals high and fraud low at the same time?

If the answer is “not yet,” that’s the work. And it’s worth doing before the next forcing function arrives.