Financial Services · AI SDR / Retail Communications

Broker-dealer achieved FINRA-compliant AI outbound — 312 principal reviews automated

Published May 20, 2026

The problem

The customer ran an AI SDR generating outbound emails to retail investors for a new product line. Their CCO flagged the workflow immediately: FINRA Rule 2210 requires principal pre-approval for retail communications, and the firm had no mechanism to meet that requirement at AI-generation volume. The options were (a) shut down the AI SDR, (b) hire 3 principals to manually review every email, or (c) find a technical solution.

The risk

FINRA Regulatory Notice 24-09 explicitly states that AI use does not relieve firms of supervisory obligations. An examination finding for unsupervised AI retail communications carries fines and heightened scrutiny. The customer's compliance counsel estimated the exposure at $200k–$500k per examination finding.

Implementation

Wired the AI SDR output through POST /v1/gate/outbound with the FINRA retail communications policy bundle. All AI-generated retail emails are routed to the principal review queue automatically — the principal receives a Slack card with the email content, the specific rule that fired (if any), and one-click approve/reject. Clean emails queue for review; emails with WARN-level violations (performance claims, superlative language, missing disclosures) are flagged with the exact violation highlighted. Every approval and rejection is logged with principal identity, CRD number, timestamp, and decision hash.

What changed

• 312 retail communications reviewed by principals in the first 60 days.
• Average review time: 47 seconds per item (down from 8 minutes for manual review).
• Zero FINRA examination findings during the subsequent regulatory review.
• Audit export produced in 90 seconds for examiner request — 312 chained log entries with approver CRD.

FINRA-compliant AI SDR — principal review gate

// Every AI-generated retail communication passes through the gate.
// The FINRA policy bundle flags: performance claims, superlatives,
// missing disclosures, guarantee language, unlicensed activity assertions.

const result = await dv.intercept({
  recipient: prospect.email,
  subject:   aiSubject,
  body_html: aiGeneratedContent,
  metadata: {
    rep_crd:            representative.crdNumber,
    firm_crd:           firm.crdNumber,
    communication_type: "retail",  // triggers mandatory principal review
    product_category:   "fixed_income",
  },
});

// result.status is always "QUEUED" for retail communications —
// principal must explicitly approve before dispatch.
// result.review_url → direct link to the approval card for the principal.