Compliance Memory

The institutional memory every agent, module, and analyst builds on. Layer 3 of the Strise platform.

Compliance Memory is the layer that turns every decision your team — or your agents — make into structured, re-usable intelligence. The next case arrives with the last case’s judgment already baked in.

The Thesis

Compliance teams make the same decisions over and over because their systems forget. A false positive dismissed in August turns up again in October with the same evidence, and someone re-does the work. An MLRO’s policy override in January doesn’t inform the automated screening in March.

Compliance Memory fixes this. Every disposition, every escalation, every override is logged with its reasoning, the data version it was made on, and the policy that applied. When the next case hits, agents and analysts read from the same memory and build on it.

What’s In Memory

What’s logged	Why it matters
Entity onboarded	The starting state — who, when, and the risk class at the time
Alert dispositions	Every PEP, sanctions, and adverse-media hit, with true-match / false-positive call, who made it, and why
Escalations	Which alerts went up to the MLRO, and how they were resolved
Policy overrides	When an MLRO accepted an exception to the standard rule, with the documented rationale
Re-review events	Whether a re-review ran fully automated, partially, or required manual touch
Data source versions	Which registry snapshot or screening list generated each decision
Applied policy version	Which version of your risk policy was live at decision time

Example: One Entity Over Time


Memory · Shady Sharks Ltd.

Mar '24   Onboarded                UBO James resolved, classified low-risk
Aug '24   Alert #2187              False positive, sanctions-adjacent name collision. Dismissed
Jan '25   Policy override          MLRO accepts 32% indirect exposure, documented rationale
Today     Re-review auto-run       4 seconds, fully automated, zero analyst touch

By the time “today’s re-review” runs, the platform already knows: the UBO was verified, the August alert was a known collision, and the MLRO accepted the 32% exposure. Nobody re-derives any of that. The re-review finishes in four seconds because the memory carries the judgment.

What Compliance Memory Enables

Faster re-reviews. An entity that hasn’t materially changed re-reviews in seconds, not hours.
Smarter screening. Known-false-positive patterns get suppressed automatically on the next run.
Defensible audits. Every decision is recoverable with its reasoning, the data it was made on, and the policy that applied.
Institutional judgment. Your team’s accumulated know-how is in the platform, not in individual analysts’ heads.

How It’s Built

Memory is structured, not free text. Every entry has:

An actor — human user, Strise agent, or the platform itself.
A target — entity, alert, or policy evaluation.
A decision — accept, reject, dismiss, escalate, override.
A rationale — free text or structured reason code.
A data snapshot — pointer to the exact registry / screening-list version used.
A policy version — which risk-policy version was live.

SuperData — the data layer memory points back to.
RiskLense — the policy layer memory records.
Monitoring System — where change detection meets memory.
Editing Entities — how human edits become memory entries.