Pi Shield

Pi Shield is PayIntelli's real-time fraud detection and prevention engine. Every transaction is evaluated by Pi Shield before it ever reaches a payment provider — and a decision is made in milliseconds: let it through, block it, or flag it for a human to review. Pi Shield is built differently. It combines a configurable rules engine for the fraud patterns you already know about with a machine learning model that learns from your actual transaction data — and improves continuously over time. Pi Shield evaluates every transaction before it is routed to a payment provider. It assembles a picture of the transaction from multiple signals, scores it for fraud risk, and takes action — all in under 50 milliseconds.

• Fraudulent transactions are stopped before they reach an acquirer, reducing chargebacks, dispute costs, and the operational burden that comes with them.
• Genuine customers are not unnecessarily blocked — the system is built to keep false positives low, so good transactions get through with minimum friction.
• Your fraud rules and risk thresholds are fully configurable, so Pi Shield operates within your specific risk appetite and business context.
• Novel fraud patterns are detected at the network level and reflected in scoring models automatically — your protection does not become outdated.
• Manual review effort is focused only on the cases that genuinely require human judgment, not on every flagged transaction.
• A full audit trail of every fraud decision is available at all times, with no manual reporting required.

Pi Shield sits at the very beginning of the payment flow — before orchestration, before authorisation — which means fraudulent transactions are identified and filtered out before they consume any processing resource or generate any cost.

1. Rule-Based Checks Deterministic controls that apply regardless of the fraud score. These are used for patterns your business already knows are high risk — specific geographies, velocity thresholds, known bad actors. Rules can be added, updated, or removed at any time without affecting the underlying model.

2. Machine Learning Model A scoring model trained on your transaction history that identifies fraud probability based on patterns across hundreds of signals simultaneously. Unlike rules, the model does not need to be told exactly what to look for — it surfaces patterns that may not be visible to a human analyst. The model is updated continuously as new outcomes are recorded.

The two layers work together. Rules catch what you already know. The model catches what you do not.

Some fraud attacks are not obvious at the transaction level. Card testing — where fraudsters run a large number of small transactions to verify stolen card details — looks innocuous on any single attempt. Pi Shield detects these patterns at the aggregate level through velocity controls.

Velocity rules monitor activity across multiple dimensions simultaneously — card number, device, IP address, and email — within configurable time windows. When a threshold is breached, transactions from that source are automatically blocked for the remainder of the window. No individual transaction needs to look suspicious for the pattern to be caught.

Transactions flagged for manual review do not land in a flat list. Each case arrives with full context — the signals that triggered the flag, the fraud score, the transaction value, and the customer's recent history. Analysts see exactly why the transaction was flagged and can make an informed decision quickly.

Every decision made in the review queue — approve or decline — is recorded as an outcome label and fed back into the scoring model. Over time, this means Pi Shield becomes increasingly accurate for your specific transaction patterns, not just a generic fraud population.

For businesses operating in Europe under PSD2, Pi Shield includes logic to identify transactions that qualify for Strong Customer Authentication exemptions through transaction risk analysis. Low-risk transactions that meet the regulatory criteria are flagged as eligible, allowing you to request exemptions from your acquirer and reduce authentication friction for genuine customers without compromising on compliance.

Every action Pi Shield takes is fully logged — score, signals, rules triggered, and decision — giving compliance teams an auditable record without any manual assembly.