What Happens When a Transaction Almost Breaks the Rules?

Rules are useful because they create a consistent way to identify transactions that meet defined conditions. A threshold is crossed, an alert is raised and the team knows that an investigation may be required.

But there is an interesting question at the boundary: what should an organisation do with activity that comes close to its monitoring rules without actually triggering them?

This is one of the reasons Decision AI is becoming a useful subject for transaction monitoring teams. The opportunity is not to discard established rules. It is to add information that helps investigators understand what the rules did not flag.

Illustrative view: rule outcomes and AI scores are separate signals; reviewers decide what warrants further attention.

Binary outcomes are clear—but incomplete

Consider a simple illustrative rule that flags transactions above a defined amount, subject to the organisation's broader monitoring logic. A transaction just below that level may not trigger the rule, while a transaction slightly above it does.

That difference is intentional. Rules require boundaries. Yet the activity on either side of a threshold may have many characteristics in common.

Looking only at the triggered/not-triggered outcome can obscure this proximity. A rule result explains whether the defined condition was met. It does not necessarily describe how similar the transaction was to activity the organisation intended to review.

A second signal alongside the rules

FraudShield's documented AI risk-scoring capability addresses this particular question. When an AI model is enabled for the active rule plan, processed transactions can receive an AI risk score from 0 to 100 alongside their rule results.

The score is designed to reflect how close the transaction came to breaching the rules in that plan. It is not a universal probability of fraud and should not be represented as one.

Importantly, the score and the triggered rule outcome are separate signals. A transaction can have a relatively high AI score without any rule having been flagged. That is the type of activity the complementary signal is particularly useful for surfacing.

What should happen next?

A high score by itself is not a final investigation conclusion. It is a reason to look more closely at the activity in the context of the organisation's monitoring policy.

An investigator might examine the transaction's underlying characteristics, previous activity and the rules that influenced the score. Depending on configuration, FraudShield also supports automatic case creation for a transaction that triggered no rules but falls within an enabled AI-score range.

This is a controlled step from a supplemental signal to a defined workflow. It does not mean the model has proven that the transaction is suspicious or that it has replaced a rules-based monitoring programme.

Why explainability matters near a threshold

An investigator needs to know why a score is elevated. Otherwise, the system creates a new queue without making the work easier to understand.

FraudShield's documentation describes rule-level explanations showing the impact and relative influence of processed rules on the AI risk score. Those details can help the reviewer understand which parts of the rule plan contributed most to the assessment.

The result is more useful when the team can see the underlying factors and decide whether further investigation is proportionate.

A practical illustration

Imagine two transactions that trigger no configured rules. One is comfortably inside the expected ranges of the rule plan. The other sits close to several monitored boundaries.

A binary rules-only view may show the same outcome for both: not flagged. A complementary AI score may help distinguish how close each came to the rule conditions.

This does not establish whether the second transaction involves financial crime. It gives the team an additional basis for deciding whether the activity deserves attention, using their own policies and investigative processes.

What should teams measure?

Before celebrating more cases, teams should ask whether the additional signal identifies useful review opportunities, how much investigation time it creates and whether the outcomes justify that effort.

The right configuration depends on the organisation's volumes, risk tolerance and monitoring requirements. A score threshold that creates too much low-value work can undermine the intended benefit.

The value at the boundary

The interesting advance is not that AI can simply declare a transaction risky. It is that a defined, explainable signal can add nuance to an otherwise binary monitoring result.

For Nexiant's FraudShield solution, this makes Decision AI a concrete operational discussion rather than a generic claim about intelligent fraud detection.

Rules tell us when a boundary has been crossed. A complementary score can help us see activity approaching that boundary—and decide whether another look is warranted.

Further reading

Read FraudShield's AI risk-scoring guide or contact Nexiant.

Frequently Asked Questions

This article was accurate at the time of publication in November 2026 and is intended for general informational purposes only. It does not constitute legal, regulatory or compliance advice. Organisations should seek qualified professional guidance in relation to their specific obligations.