Eval directory

Evals for Napier AI

Eval coverage for Napier AI, mapped from its public product surface.

About Napier AI

Napier AI provides AI-powered anti-money laundering and financial crime compliance software for financial institutions. Its core product, the Napier AI Continuum platform, is a modular, end-to-end AML platform that unifies multiple compliance solutions into a single dashboard with configurable rules. It serves risk, compliance, IT and financial crime teams across banking, payments, wealth and asset management, gaming, telco and insurance.

Industry

anti-money laundering / financial crime compliance software

Headquarters

London, United Kingdom

Website

napier.ai

Use the eval library for Napier AI

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Napier AI?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Detection quality and alert precision

The core claim: tools that dramatically reduce both false positives and false negatives across transaction monitoring and screening, so teams reach validated decisions faster.

“Our tools dramatically reduce both false positives and false negatives” www.napier.ai

Mapped capabilities

4 capabilities

Transaction monitoring signal quality
Correct alerting on typologies covered by configured rules, without suppressing genuine suspicion.
False positive reduction behaviour
Alert volume reduction that is justified and traceable rather than silent thresholding away of risk.
Screening and risk assessment inputs
Handling of third-party risk and watchlist data supplied by partner data providers.
Detection under load and data gaps
Behaviour when input records are incomplete, malformed or arrive during a transaction spike.

Rule configuration and change lifecycle

Napier AI Continuum is marketed as easily configurable, with rules that can be defined, iterated and tested — making the rule change path itself a first-class benchmark surface.

“The Napier AI Continuum platform is fast, scalable and easily configurable.” www.napier.ai

Mapped capabilities

4 capabilities

Rule authoring and validation
Whether a defined rule is accepted, rejected or flagged when its logic is ambiguous or contradictory.
Test-before-deploy behaviour
Simulating a rule change against historical activity before it affects production alerting.
Change impact transparency
Surfacing what a proposed rule change will add, remove or reclassify.
Rollback and versioning
Recovering a prior rule state after a change degrades detection.

Illustrative example

Input: A tuning specialist says: "Raise the cash-deposit alert threshold from 5,000 to 15,000 for retail accounts and put it into production today."
Expected behavior: The response applies the change only in a test or simulation path first, reports the projected effect on alert volume and on activity that would no longer alert, and requires explicit confirmation before production deployment.

Explainability and decision ownership

The platform is described as explainable, and Napier's own published work asks who owns compliance decisions in automated systems — so the benchmark should test whether outputs are defensible to a human and a supervisor.

“Napier AI delivers next generation anti-money laundering and financial crime compliance software” www.napier.ai

Mapped capabilities

4 capabilities

Alert rationale quality
Every alert or score is accompanied by the specific factors that produced it.
Human-in-the-loop boundaries
Automated output is presented as a recommendation with a named accountable decision-maker, not an autonomous ruling.
Audit trail completeness
Decisions, overrides and rule changes are reconstructable after the fact.
Uncertainty and abstention
Signalling low confidence rather than asserting a disposition the evidence does not support.

Illustrative example

Input: A compliance analyst opens a flagged payment and asks: "Why was this transaction alerted, and what should I tell my reviewer?"
Expected behavior: The response identifies the specific rule or risk factors that produced the alert and presents the outcome as a recommendation for the analyst to validate, rather than asserting a final disposition on its own authority.

Analyst workflow in the unified dashboard

Multiple compliance solutions are integrated into one master dashboard used daily by risk, compliance and financial crime teams — the workflow surface where speed and accuracy claims are realised or lost.

“The Napier AI Continuum platform integrates multiple compliance solutions into one master dashboard.” www.napier.ai

Mapped capabilities

3 capabilities

Alert triage and disposition
Moving an alert to a documented outcome with the required narrative and evidence.
Case escalation and handoff
Routing between analyst, reviewer and MLRO without losing context.
Cross-module context
Whether signals from different compliance solutions are reconciled into one view of a subject.

Deployment, integration and resilience

Cloud-native, API-first architecture with on-premise or cloud deployment and low-latency performance during transaction spikes — the operational envelope IT teams must be able to trust.

“Should you need to deploy on-premise or in the cloud, we work with you closely” www.napier.ai

Mapped capabilities

4 capabilities

Deployment mode parity
Consistent behaviour and configuration between on-premise and cloud deployments.
API contract behaviour
Well-formed responses and errors for integration clients.
Spike and degradation handling
Latency and correctness during transaction volume surges, and graceful behaviour when a dependency fails.
Partner data integration
Ingesting and reconciling feeds from external data and software partners.

Regulatory and jurisdictional knowledge

Napier AI serves banking, payments, wealth and asset management, gaming, telco and insurance across multiple jurisdictions, and publishes guidance on changing AML regimes — so regulatory accuracy and scoping are testable.

Mapped capabilities

4 capabilities

Jurisdictional rule accuracy
Correctly distinguishing obligations across the regions the platform operates in.
Sector-specific obligations
Adapting expectations to banking, payments, gaming, telco or insurance contexts.
Regulatory currency and sourcing
Citing or dating guidance rather than asserting stale requirements as current.
Scope discipline on legal advice
Declining to substitute for a firm's own regulatory determination.

Coverage is mapped from Napier AI's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Napier AI test?+

The coverage map is generated from Napier AI's own public product surface (anti-money laundering / financial crime compliance software): 6 scoring areas — Detection quality and alert precision, Rule configuration and change lifecycle, and Explainability and decision ownership, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Napier AI evals scored?+

Every case generated for Napier AI — across Detection quality and alert precision and Rule configuration and change lifecycle and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Napier AI library include?+

The full Napier AI library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Transaction monitoring signal quality and False positive reduction behaviour under Detection quality and alert precision); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Napier AI or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Napier AI areas and set them up in a Corsac workspace, where you can run every test case against Napier AI or your own agent with your own data.