Engine 1 · TRE
Trust & Risk Engine
The AI-assisted governance layer that converts every sustainability claim in fund factsheets, websites, KIDs, and SFDR disclosures into structured, evidence-linked, human-approved, supervisor-ready records.
IN SHORTThe Dollarbaz Trust & Risk Engine converts every sustainability claim into structured, evidence-linked, human-approved, supervisor-ready records. Deterministic scoring. Human in the loop. Immutable audit trail.
Core capabilities
Document ingestion
Paste, PDF upload, or live URL. The engine reads the full document — not just a snippet.
25 claim types (CT01–CT25)
Standardised taxonomy classifies each claim. Evidence checklists generated per claim from pre-defined rules.
Deterministic risk score 0–100
Flag-based logic, fully explainable. No AI in the scoring path. Every flag has a defined rule. Every score is reproducible.
Structured review & approval
Full decision logging with timestamp, reviewer identity, evidence state, and claim version at time of approval.
Evidence drift monitoring
Alerts when evidence expires, data vendor methodologies update, or holdings changes render claims technically unsupported.
Immutable audit trail
Hash-chained, append-only log. Every state change is cryptographically verifiable. No record can be deleted or altered.
Audit pack export
On-demand, supervisor-ready export. Ready for FCA supervisory requests the moment they arrive.
Monitoring dashboard
Evidence completeness %, overdue claims, approval rate, and risk score distribution across your portfolio.
Claim & evidence taxonomy
25 claim types (CT01–CT25) and 25 evidence types (ET01–ET25). Every CT code has a corresponding ET rule matrix — evidence requirements are deterministic, not discretionary.
Claim types — sample
Evidence types — sample
Deterministic scoring engine
Risk score (0–100) calculated from a defined flag matrix. Every flag has a fixed weight. Every weight has a documented rationale. Run the same inputs — get the same output. Explainable to a regulator, line by line.
No AI in the scoring path.
Architecture principle
AI assists extraction, classification, and summarisation. Rules and humans decide. Every AI action is explainable and logged.
Human approval is required for every governed action. The scoring engine is deterministic and rule-based — humans control every decision.
Claim lifecycle
Every claim moves through a defined state machine. Transitions are logged immutably. No state can be skipped. No record can be deleted or altered.
Draft
Claim extracted or entered. AI suggests classification. Human reviews.
Under Review
Evidence checklist generated. Reviewer attaches evidence per ET code.
Approved
Decision logged with evidence state at time of approval. Immutable.
Rejected
Rejection reason required. Full decision chain preserved. Rework routed.
Monitoring
Approved claims enter drift detection. Alerts on evidence expiry.
SFDR 2.0 & ESMA upgrades — built in
Regulatory readiness is not a future release. These guardrails are live in the current engine.
SFDR 2.0 Label-Agnostic Mapping
When a claim is classified, the engine auto-suggests how it maps to proposed SFDR 2.0 product categories and whether supporting evidence meets minimum sustainability thresholds.
ESMA Thematic Note Guardrails
Pre-built rule checks for ESG integration and exclusion claims per the January 2026 ESMA thematic note. Automatically flags vague language that ESMA has specifically targeted.
FCA Drift Detection
Evidence expiry alerts tied directly to data-vendor refresh cycles and holdings changes. Catches undetected drift in real time — the FCA requires claims to remain accurate and current.
Start governing claims
in 30 days.
30-day pilot. 1–3 funds. Concierge onboarding. FCA-ready audit trail from day one.