Customer case study

Your CDD policies are only as good as how consistently they are applied

How AI agents are closing the gap between written policy and daily compliance execution under pressure.

Profile

  • Industry: Investment management
  • Location: Singapore (MAS-regulated)
  • Entity type: Licensed fund manager
  • Key challenge: Inconsistent CDD policy application
  • Solution: Prudexis AI Compliance Agent
  • Outcome: Policy-grounded reviews with a full audit trail

The real problem with CDD

Most compliance teams already have strong CDD policies. They define escalation triggers, risk classifications, source-of-wealth expectations, and documentation standards.

The issue is consistency. Under workload pressure, analysts apply interpretation differently, and narratives start reflecting reviewer style instead of firm policy.

This is not a people problem. It is a systems problem. Prudexis addresses it by grounding agent reasoning directly in each firm's uploaded policy framework.

The insight

"Your CDD policies are only as good as how consistently they are applied. With the agent, consistency depends on policy, not reviewer variability."

Prudexis does not apply generic AML logic. The agent uses the firm's own definitions of risk, materiality, escalation criteria, and documentation expectations.

Every review, rationale, and recommendation is prepared against that policy framework and then confirmed by a human analyst.

The workflow

Agent prepares.

For periodic or event-driven review, the agent evaluates profile changes, activity, screening results, and known risk factors, then drafts the full CDD narrative.

Dashboard surfaces the work.

Analysts open a pre-prepared queue with cases ready for approval, edit, or escalation.

Analyst reviews and confirms.

The analyst applies judgment and signs off. Agent reasoning and analyst decision are recorded in one audit trail.

Why this is different from template automation

Scheduled workflows and templates solve timing and formatting. They do not solve reasoning quality.

A policy-grounded agent can explain risk classification in the firm's own terms, assess whether unusual activity remains policy-compliant, and document why no escalation was required.

That creates a defensible answer for regulators: policy was systematically applied, and the analyst reviewed and confirmed the outcome.

Example output

PRUDEXIS AGENT - PERIODIC CDD REVIEW

PRUDEXIS AGENT

Wei Lin Chong · ID: 1000000041 · Annual refresh

Risk rating:

✓ Maintain Medium

- no upgrade or downgrade criteria met under firm policy.

Flags: One item noted and assessed. Account activity increased approximately 18% year-on-year in Q3 2025. Determined to be within acceptable range for this customer's income profile. Documented per firm policy on activity variations. No further action required.

Prepared narrative for analyst approval:

Annual periodic review completed for Wei Lin Chong (ID: 1000000041). Profile reviewed against current KYC records, watchlist and adverse media results, and 12 months of transaction activity. No material changes identified. Source of wealth assessed as consistent with disclosed profile per firm CDD policy. A Q3 2025 activity increase of approximately 18% year-on-year was noted, assessed, and determined to be within the expected range for this customer's income profile, documented in accordance with firm policy on activity variations. Risk classification maintained at Medium. Cross-border activity assessed as consistent with stated occupation and within firm policy parameters. No escalation recommended. Next scheduled review: April 2027.

Status: Ready for approval. Agent reasoning and analyst decision captured in audit trail.

The analyst confirms the narrative in under two minutes, with both machine reasoning and human decision captured for audit.

The results

  • Consistent policy execution: Reviews follow firm CDD criteria case by case
  • Audit-ready documentation: Narratives show why decisions were made
  • Edge-case visibility: Non-standard profile shifts are surfaced explicitly
  • Operational reliability: Review cycles are completed on schedule
  • Stable quality: Output quality does not vary by analyst seniority or workload spikes

Customer quote

"The policies were always there. Now they are actually being applied to every customer, every time."

Head of Compliance, MAS-licensed fund manager, Singapore