CHAPS Direct Access: The AI Case for Going Direct

Sterling high-value payments are having a quiet “reset moment.” The UK’s renewed Real-Time Gross Settlement (RTGS) service (RT2) is live, and it’s doing something more important than a technology refresh: it’s changing what “good” looks like for CHAPS participation.

If you’re a bank, building society, or regulated PSP still using indirect CHAPS access through a sponsor, the old rationale—“direct is expensive and complex”—is getting weaker. Not because the work disappears, but because the risk profile of staying indirect is getting harder to defend, and the upside of direct participation is getting easier to quantify.

This post is part of our AI in Payments & Fintech Infrastructure series, so I’m going to take a clear stance: RT2 makes CHAPS direct participation not just a payments strategy decision, but an AI and data strategy decision. The institutions that treat it that way will move faster, manage liquidity better, and spot fraud and scams earlier.

RT2 changes the CHAPS conversation: control becomes the product

RT2’s headline goals—greater access, increased resilience, wider interoperability, and improved user functionality—sound like infrastructure language. The real impact is commercial: CHAPS becomes a platform you can actively optimize, not a rail you passively consume.

Indirect CHAPS access was a sensible stepping-stone for years. But the trade-off has always been the same: you outsource control in exchange for lower upfront cost and quicker entry.

With RT2 live, that trade-off is sharper because high-value payments aren’t “set-and-forget” anymore. Boards expect:

• Stronger operational resilience (and evidence you can run through failures)
• Better intraday liquidity control (and fewer surprises at 4:45pm)
• Better data (ISO 20022 is now a competitive ingredient, not a compliance afterthought)
• Faster change cycles (new products, new hours, new reporting demands)

Control is the product because control determines customer experience, cost-to-serve, and your ability to innovate.

Why indirect CHAPS access is becoming a strategic constraint

Indirect access doesn’t just add a party to the process. It changes how you operate.

1) You inherit your sponsor’s priorities and failure modes

When you rely on a sponsor bank, your CHAPS service quality is bounded by:

• Their operational windows and cut-off interpretations
• Their routing preferences and risk appetite
• Their incident management and recovery playbooks
• Their change calendar (which may not match yours)

Even if your sponsor is excellent, you’re still accepting a structural limit: you can’t fully design your own resilience. For regulated firms, that’s not a small point—especially as expectations around third-party dependencies and operational resilience tighten.

2) Liquidity management stays “approximate”

High-value payments are liquidity management with customer-facing consequences. Indirect participation often means:

• Less precision in intraday forecasting
• Less flexibility in timing and queue management
• More operational choreography between teams and providers

That creates a familiar pattern: operations builds buffers “just in case,” treasury compensates with conservative positions, and everyone pays for it.

3) ISO 20022 data is less useful when you can’t fully act on it

ISO 20022 brings richer, more structured payment data. But value comes from what you can do with that data in real time:

• Automated exception handling
• Better sanctions/AML triage using contextual fields
• Cleaner reconciliation and fewer repair cycles
• Smarter fraud and scam detection using behavioral signals

If your model forces you through sponsor workflows, you often end up with partial visibility and delayed control—exactly the wrong combination for AI-driven operations.

Direct CHAPS participation: what you gain (beyond the obvious)

Direct participation means connecting directly to the Bank of England’s RTGS for settlement. The “obvious” benefits are real: better control, stronger independence, and the ability to manage your CHAPS proposition without sponsor constraints.

But the more interesting gains show up when you connect direct participation to AI-enabled payments infrastructure.

1) AI-optimized routing and queue management becomes practical

High-value payment processing isn’t only about pushing messages. It’s about prioritizing, scheduling, and funding payments while staying within risk and liquidity constraints.

Once you’re direct, you can apply AI/ML in a way that’s hard to replicate through indirect arrangements:

• Intraday liquidity prediction using historical flows, seasonality, and customer patterns
• Queue prioritization recommendations that balance SLA commitments, cost of liquidity, and settlement urgency
• Exception prediction (which payments are likely to fail sanctions screening, format validation, or downstream rules)

One practical example I’ve seen work: building an “operations co-pilot” that flags payments likely to hit exceptions before they enter the critical path, so teams stop firefighting and start preventing.

2) Extended hours are only valuable if your ops model can keep up

RT2 enables the industry’s push toward extended operating hours. That’s good news—until you realize many firms are still running human-heavy workflows for high-value payments.

AI helps here in a grounded, non-hype way:

• Automated enrichment and validation for ISO 20022 fields
• Intelligent case assignment for investigations (sanctions/AML/fraud)
• Alert quality improvements (fewer false positives, faster decisions)

Extended hours without smarter operations just means extended stress. Extended hours with AI-assisted operations means you can deliver premium service without doubling headcount.

3) Richer data turns compliance into measurable performance

RT2 arrives at the same time as broader market momentum around ISO 20022 enhanced data, including increasing use of:

• Purpose codes
• Legal Entity Identifiers (LEIs)
• Better counterparty and remittance structuring

Many institutions treat these as compliance fields. The better approach: treat them as performance inputs.

When you standardize and validate enriched data, you get:

• Higher straight-through processing (STP)
• Faster investigations (because context is present)
• Better auditability (who paid whom, why, and under what contractual context)

AI thrives on structured data. ISO 20022 is the data layer that makes AI in payments less guesswork and more engineering.

Fraud, scams, and high-value payments: why “later” is too late

High-value rails attract high-intent crime. In late 2025, most payments leaders I speak to aren’t worried about one single fraud typology—they’re worried about how fast typologies evolve, especially where social engineering, mule accounts, and synthetic identities intersect.

Direct participation doesn’t magically stop fraud. But it improves your ability to detect and respond because:

• You can correlate CHAPS events with internal signals (login anomalies, device changes, payee changes)
• You can act faster on holds, investigations, and escalation
• You can design your control framework, not inherit a sponsor’s defaults

What AI can do specifically in CHAPS-like environments

For high-value payments, the goal isn’t “catch everything.” The goal is reduce loss without destroying the customer experience.

AI-driven fraud controls that fit CHAPS operations tend to focus on:

• Behavioral anomaly detection: unusual beneficiary changes, out-of-pattern payment timing, atypical amounts relative to entity size
• Network analytics: mapping relationships between originators, beneficiaries, and intermediaries to spot mule clusters
• Real-time risk scoring: combining payment message attributes (ISO 20022 fields) with channel telemetry
• Explainable alerts: giving ops teams the “why” so decisions are fast and defensible

A strong pattern is combining rules (what compliance requires) with ML (what behavior suggests). Rules keep you compliant; ML keeps you adaptive.

A useful internal mantra: “Speed is a control.” If you can’t decide quickly, you’re effectively opting into risk.

A decision framework: should you go direct on CHAPS now?

RT2 lowers barriers to entry and streamlines onboarding. That doesn’t mean every institution should switch immediately. It means you can evaluate it with a clearer business case.

Here’s a practical framework I’d use in an executive discussion.

1) If you sell premium treasury services, indirect access will show

If your clients care about certainty—property completions, corporate deal settlements, margin calls—then CHAPS performance becomes part of your brand. Sponsors can deliver good service, but you’re still selling someone else’s operational model.

Rule of thumb: if CHAPS is customer-visible, direct participation is strategically clean.

2) If you’re investing in AI, indirect access limits the ROI

AI programs in payments succeed when models can influence real decisions: routing, prioritization, investigations, and liquidity.

If your architecture can’t apply AI recommendations without sponsor friction, you’ll end up with “insights dashboards” instead of operational impact.

Rule of thumb: if you’re funding AI for payments ops, direct participation strengthens the value chain.

3) If you’re serious about resilience, dependency needs a hard look

Operational resilience isn’t a policy document; it’s a set of tested capabilities. Indirect access adds a critical dependency you can’t fully control.

Rule of thumb: if your tolerance for third-party concentration risk is shrinking, direct participation reduces structural exposure.

What payment leaders should do in the next 90 days

Direct participation is a program, not a purchase. Teams move faster when they treat it like a product launch with governance, data, and operating model baked in.

Here’s a focused 90-day plan that works in practice:

1. Map your current CHAPS journey end-to-end

   • Where do delays happen?
   • Where do repairs happen?
   • Which steps are manual, and why?

2. Quantify the “indirect tax”

   • Sponsor fees (explicit)
   • Operational overhead (people time, reconciliations, investigations)
   • Liquidity buffers (cost of conservative positioning)
   • Incident impact (customer remediation and reputational cost)

3. Define the AI opportunities in plain language

   • “Predict intraday liquidity needs with X% error band”
   • “Reduce repair rate by Y% by pre-validation”
   • “Cut investigation time from N hours to M minutes for top scenarios”

4. Design the operating model for extended hours

   • Decide what’s automated vs. supervised
   • Set alert thresholds and escalation policies
   • Align fraud, AML, and payments operations on one playbook

5. Create a direct participation readiness checklist

   • ISO 20022 data quality controls
   • Monitoring and incident response
   • Business continuity and contingency testing
   • Internal SLAs for investigations and approvals

If you do only one thing: treat data quality as a first-class deliverable. AI can’t compensate for missing or inconsistent ISO 20022 fields—it just makes the errors faster.

Where RT2 fits in the bigger AI-in-infrastructure story

RT2 is a reminder that financial infrastructure upgrades create rare windows where competitive positions can shift. When the foundation changes, the teams that modernize their operating model—not just their connectivity—pull away.

Direct CHAPS participation isn’t mandatory for everyone, but the direction of travel is clear: more access, more interoperability, richer data, and higher expectations for resilience. AI is the amplifier that turns those capabilities into measurable outcomes.

If you’re weighing whether to stay indirect or go direct, don’t frame it as “cost vs. complexity.” Frame it as:

• How much control do we need to deliver the service we promise?
• How quickly can we detect and stop high-value fraud and scams?
• Can we use ISO 20022 data to reduce cost-to-serve, not just pass audits?

The next wave of payments leaders won’t be defined by who connects to the rails. They’ll be defined by who can operate the rails intelligently.

What would your CHAPS service look like if you could tune liquidity, routing, and fraud controls in near real time—without waiting on a sponsor bank to change their roadmap?