- Rising fraud losses trace to a decisioning gap, not a detection gap: fraud models are accurate, but most FS architectures have no layer that turns a score into an enforced decision while the transaction is still open.
- US consumer fraud losses hit $15.9 billion in 2025, up roughly 430% since 2020, with the true cost closer to $196 billion once underreporting is counted.
- A late decision costs twice: fraud that gets through, plus an estimated $81 billion a year in false declines from legitimate transactions blocked on stale context.
- Most fraud stacks split state, scoring, and enforcement across three separate systems, creating race conditions fraudsters exploit and no single audit trail of the decision.
- Instant payment rails (SEPA Instant, FedNow, UPI) settle in seconds and don’t reverse, turning batch decisioning from a performance problem into a compliance one.
- A decisioning layer, not a faster model, is what holds state, runs logic, enforces the outcome, and records the trail in one transaction, and it’s also the foundation that makes AI trustworthy in production fraud prevention.
TL;DR
The gap nobody budgets for
Financial services has never been better at detecting fraud. Models are accurate, scores arrive in milliseconds, and AI investment keeps climbing. Yet losses climb with it. That is not a detection problem. It is a decisioning problem. As AI models arrive inside these complex, data-driven infrastructures, high-performance decision engines are being paired with fragmented, often stale data foundations, and the results show.
A fraud score computed after the authorization window closes is an audit record, not prevention. The model was right. The alert fired. The money moved anyway. Between detection and enforcement, most FS architectures have a gap: nothing turns a score into an enforceable, auditable decision while the transaction is still waiting.
What the numbers say
In testimony to the US Congress Joint Economic Committee, the FTC reported $15.9 billion in consumer fraud losses for 2025, up roughly 430% since 2020. Accounting for underreporting, the true annual cost may be closer to $196 billion. Nearly two thirds of victims send money within 24 hours of first contact, increasingly over rails where settlement is instant and irreversible.
And a late decision costs you twice. Alongside the fraud that gets through, false declines cost North American merchants and issuers an estimated $81 billion a year and climbing: legitimate customers turned away by decisions made on stale context. The loss is not only the transaction. It is often the customer.
INFOGRAPHIC
The architecture problem
Three things must happen simultaneously for a real-time decision to hold: the current state must be read, the logic must run against it, and the outcome must be enforced. In most stacks those three live in three different systems: one holds the state, an ML service produces the score, and a message queue carries the block instruction.
Concurrent checks read the same velocity counters before any of them are updated. That race condition is exactly what fraudsters exploit with rapid parallel transactions. And when the pieces are separate, there is no single audit trail of the decision sequence: what was queried, how current it was, what was scored, and what was decided.
The regulatory clock is running
- UK APP reimbursement rules put prevention failure directly on bank P&L.
- SEPA Instant is mandatory; FedNow adoption has passed 1,400 institutions.
- ISO 20022 migration is complete on FedWire and CHIPS: richer data, same shrinking window.
- RBI now requires real-time screening across UPI, at 18B+ transactions a month.
Instant rails settle in seconds and do not reverse. Once funds release, they are unrecoverable. Every step, validating, assessing, approving or declining, settling, releasing, has to complete inside that sub-second window. Batch decisioning is now both a performance drag and a compliance risk, and the risk compounds as high-throughput AI models are bolted on top of it. The issue is not the AI. It is the data infrastructure underneath it.
The decisioning layer, defined
A decisioning layer holds authoritative state, runs the decision logic against it, enforces the outcome, and records the full decision trail, all within one transaction, while the payment is still waiting. Not faster plumbing. A layer with decision authority.
It is also the foundation any AI ambition depends on: as our Field CTO Dheeraj Remella argues, most of what makes AI trustworthy in fraud prevention is the deterministic infrastructure underneath it, not the model itself.
Where to go next
For the business case, read Ken Ballou’s executive brief, Why False Declines Cost More Than Fraud in AI-Driven Payments. For the hands-on view, the Volt Dev Hub lets your architects see a real-time decision made in practice. Our CEO, David Flower, takes the FTC numbers head-on in a new piece publishing soon!
Learn more at voltactivedata.com/bfsi.
What is the decisioning gap in financial services fraud prevention?
The decisioning gap is the space between an accurate fraud score and an enforced decision, the missing step where a score should become an allow, block, or route action before the transaction settles. Financial institutions have solved detection: models are accurate and scores arrive in milliseconds. What they haven’t solved is turning that score into an authoritative decision inside the same window the transaction is still open.
Why do fraud losses keep rising even though detection technology keeps improving?
Detection and decisioning are two different problems, and most institutions have only closed the first gap. The FTC reported $15.9 billion in US consumer fraud losses for 2025, up roughly 430% since 2020, even as fraud models have gotten more accurate. A model that scores a transaction correctly after the authorization window closes has produced an audit record, not a prevented loss.
What are false declines, and why do they matter as much as fraud?
A false decline is a legitimate transaction blocked because a decision was made on stale or incomplete context, and it costs North American merchants and issuers an estimated $81 billion a year. A late or fragmented decisioning architecture produces both failure modes at once: fraud that gets through and legitimate customers turned away, which is why a late decision costs a financial institution twice.
Why does fraud architecture break down even when each individual system works?
Most fraud stacks split state, scoring, and enforcement across three separate systems: one holds account state, an ML service produces the score, and a message queue carries the block instruction. Concurrent transactions can read the same velocity counters before any of them update, which is the exact race condition fraudsters exploit with rapid parallel transactions, and no single system captures a complete audit trail across all three.
What regulations are forcing real-time fraud decisioning right now?
Several already have direct P&L or compliance consequences. UK APP reimbursement rules put prevention failure directly on bank balance sheets, SEPA Instant and FedNow are already mandatory and past 1,400 adopting institutions respectively, ISO 20022 migration is complete on FedWire and CHIPS, and India’s RBI requires real-time screening across UPI at more than 18 billion transactions a month. Instant rails settle in seconds and do not reverse, which turns batch decisioning into a compliance risk, not just a performance one.
What is a decisioning layer, and how is it different from a fraud model?
A fraud model produces a score. A decisioning layer holds authoritative state, runs decision logic against it, enforces the outcome, and records the full decision trail, all inside one transaction while the payment is still waiting. It’s the layer that turns a model’s output into an enforceable, auditable action rather than faster plumbing for the same score.
Does better AI fix the fraud decisioning problem on its own?
Not without a deterministic layer underneath it. AI models can keep improving detection while losses keep climbing, because the constraint isn’t model quality, it’s whether the infrastructure underneath can turn a model’s output into a decision in time. What makes AI trustworthy in fraud prevention is the deterministic infrastructure it runs on, not the model itself.
About Author
Featured Resources
