Trovix AI Integration

Trovix Audit ©

AI governance and audit trail layer for all AI systems in your firm

Audit is the governance infrastructure that sits across every AI system operating in your firm — Trovix products and any third-party AI tools you use. It captures every AI-assisted decision, the inputs used, the model version, the confidence score, the human reviewer action and the rationale for any override. The resulting log is tamper-evident, timestamped and structured to satisfy the documentation requirements of FCA Consumer Duty, PRA SS1/23, SRA supervision obligations, FRC audit guidance and ICO accountability requirements. Audit also generates the model performance reports and governance documentation required by your risk and compliance function.

Systems Audit integrates with

Audit collects decision events from any system that exposes an API or can be instrumented with a logging agent. It is not limited to Trovix products.

All Trovix products
Native integration. Every Trovix AI decision event automatically sent to Audit with full metadata.
Microsoft Azure OpenAI
API-level logging agent captures every prompt and completion. Logs include deployment name, model version and token counts.
Anthropic Claude API
API logging agent captures requests and responses. Integrates with Anthropic usage metadata for model version tracking.
Third-party AI tools
Where third-party tools expose webhooks or API events, Audit can consume and log decision events. Covers Harvey, Luminance and similar.
Power Automate / Logic Apps
AI steps in Power Automate flows instrumented to send decision events to Audit. Covers AI Builder and Azure AI Services calls.
Custom ML models
Logging SDK available for Python and Node.js. Any ML model can be instrumented to send prediction events with input features and output scores.

Integration architecture

AI Decisionany Trovix productHuman Actionreview / overrideAudit Engineimmutable log layerTimestamp + InputModel Version + ScoreHuman RationaleAudit Storetamper-evident logRegulator ExportFCA / SRA / FRCInternal Dashboardreal-time viewEvery log entry is cryptographically timestamped. Override rationale is mandatory before any AI decision can be countermanded.

Technical specification

Log capture
Event-driven capture via API, webhook or logging agent. Each event includes: timestamp, system ID, decision type, inputs, output, confidence, model version.
Immutability
Log entries are cryptographically hashed on creation. Any tampering is detectable. Hash chain maintained across the log sequence.
Human review capture
Override events capture: reviewer identity, action taken, rationale text and timestamp. Mandatory rationale for any decision countermand.
Retention
Configurable retention period. Default 7 years for regulated firm compliance. Immutable archive export to AWS S3 or Azure Blob Storage.
Reporting
Pre-built reports: model accuracy over time, override rate by decision type, reviewer activity, confidence distribution. Exportable to PDF and CSV.
Access control
Role-based access. Compliance function sees all logs. Business users see logs for their own decisions only.
Deployment
Deployed inside your cloud tenant or on-premise. Log data never leaves your environment.

Regulatory compliance

FCA Consumer Duty PS22/9AI-assisted customer decisions must be explainable and auditable. Audit provides per-decision explanation and confidence score for regulatory review.
PRA SS1/23 Model RiskModel risk management requires documented model validation, performance monitoring, version control and override logging. Audit covers all four.
SRA Code r.2/r.4Supervision obligation requires evidence of partner oversight of AI-assisted work. Audit provides timestamped review and sign-off records.
FRC ISA (UK) 200FRC guidance on AI in audit requires documentation of AI involvement and professional judgement. Audit generates per-procedure AI involvement log.
ICO Accountability Art.5(2)Controllers must demonstrate GDPR compliance. Audit provides evidence of purpose limitation, decision logging and human oversight of automated processing.
ISO 42001AI management system standard requires documented AI governance processes. Audit provides the technical governance layer for ISO 42001 certification.

Discuss Audit deployment across your AI systems

A technical conversation about your existing systems and integration approach. 30 minutes.

Book a technical call