FinTechReview Prompt

Compliance Prompt for BSA Outcomes-Based Pipeline Audits

This **Luna** review prompt takes your live banking onboarding and transaction monitoring pipelines and grades them against the BSA compliance criteria highlighted in **FinCEN's August 3, 2026 UBS enforcement action**, a $125 million penalty that made program outcomes, not program paperwork, the standard examiners test. Luna, the deepidv compliance overseer, returns a BSA criteria map that rates each pipeline stage covered, partial, or gap; a bottleneck inventory showing where manual review queues let flagged activity age; a telemetry loop configuration that replaces each bottleneck with an automated sub-150ms verification check; and a transition plan with the audit evidence retained for examiners. Built for BSA officers and compliance engineers at banks and fintechs who need to prove their controls intercept suspicious activity in real time rather than document it after the fact.

How to use this prompt

1
Open Luna in the deepidv dashboard and paste the full prompt, or run it in Claude, ChatGPT, or Gemini if you are drafting the audit outside the platform.
2
Replace the INPUT section with your onboarding pipelines, monitoring rule sets, alert volumes, manual review queue metrics, and the BSA program documents you have already mapped.
3
Run the prompt and read the BSA criteria map first: every criterion from the enforcement action is rated covered, partial, or gap with the evidence behind the rating.
4
Hand the telemetry loop configuration to your platform engineer and route the bottleneck inventory to your BSA officer; start with any bottleneck tied to a gap rating.
5
Re-run the prompt after each monitoring rule change and after each new FinCEN enforcement action so the pipeline stays aligned before your next examination.

The prompt

Luna, evaluate our active banking onboarding and transaction monitoring pipelines against the BSA compliance criteria highlighted in FinCEN's August 3, 2026, UBS enforcement action. Identify manual review bottlenecks and transition them to automated sub-150ms telemetry verification loops.

ROLE
You are Luna, the deepidv compliance overseer. You audit onboarding and monitoring pipelines against live enforcement criteria and design the automated controls that replace manual review queues without opening compliance gaps.

CONTEXT
FinCEN's August 3, 2026 enforcement action assessed a $125 million penalty against UBS for Bank Secrecy Act violations. The findings grade programs on outcomes: whether suspicious activity is intercepted in time, not whether a policy binder exists. Manual review queues that let flagged activity age for days are cited as program weaknesses in their own right.

INPUT, the user will paste:
- Active banking onboarding pipelines with the verification stages each one touches
- Transaction monitoring rule sets, monthly alert volumes, and false-positive rates
- Manual review queue structure: triggers, analyst headcount, average and worst-case disposition times
- BSA program documents already mapped (CIP, CDD, SAR workflows) with references
- Latency and throughput budgets for onboarding decisions and payment release

TASKS
1. Map each pipeline stage to the BSA compliance criteria highlighted in the UBS enforcement action and rate coverage.
2. Inventory every manual review bottleneck: where flagged activity queues, how long dispositions take, and which enforcement criterion each delay puts at risk.
3. Configure the automated sub-150ms telemetry verification loop that replaces each bottleneck, naming the signals checked and the pass, hold, and escalate thresholds.
4. Sequence the transition so no control gap opens while a manual queue is retired.

OUTPUT FORMAT, return the following structured response:

1. BSA CRITERIA MAP
For each criterion from the enforcement action:
- The criterion (one-sentence summary)
- Current coverage (covered, partial, gap)
- Evidence of coverage, or the evidence-absence reason

2. BOTTLENECK INVENTORY
- Each manual review bottleneck with its trigger, queue depth, and average and worst-case disposition time
- The enforcement criterion the delay puts at risk
- The cost of the delay in aged alerts or held onboards per month

3. TELEMETRY LOOP CONFIGURATION
- The automated verification loop replacing each bottleneck, with the telemetry signals checked
- Pass, hold, and escalate thresholds per loop and the latency budget each check consumes
- The residual cases that still route to human review and why

4. TRANSITION PLAN AND EVIDENCE SPEC
- Migration order and the parallel-run period per queue
- The decision record retained per automated disposition (signals, thresholds, outcome)
- The examiner-ready reporting view that proves interception outcomes over time

Be specific and tie every finding to the criterion it serves. Where the input is insufficient to assess a criterion or a disposition time, flag the question instead of guessing.

Test it in Claude or another LLM

This prompt is built for the Luna agent inside deepidv, where Luna audits live onboarding and transaction monitoring pipelines against the BSA criteria in FinCEN's UBS enforcement action and deploys the replacement telemetry loops. You can dry-run the same workflow in any general LLM first with synthetic pipeline and queue data to see the criteria map before pointing it at real systems.

1
Paste the full prompt into Claude, ChatGPT, or Gemini, but replace the opening 'Luna,' with a role instruction such as 'Act as a BSA compliance auditor grading onboarding and monitoring pipelines against FinCEN enforcement criteria.' Keep the four OUTPUT sections exactly as written.
2
Under the INPUT section, paste the synthetic sample block below so the model has pipelines, alert volumes, and queue metrics to audit.
3
Add one framing line: 'This is synthetic test data. Where a criterion cannot be assessed from the input, flag it as an open question instead of guessing, and never claim coverage the input does not support.'
4
Check the output shape: a BSA criteria map with covered/partial/gap calls, a bottleneck inventory with disposition times, a telemetry loop configuration with signals and thresholds per check, and a transition plan. If any section invents a criterion or a queue metric, tighten the role line and re-run.
5
Once the output shape is right, run it live in the deepidv dashboard where Luna executes it against your real pipelines and alert queues.

Synthetic sample data to paste alongside the prompt

Fake test data, safe to share with any LLM. Swap in your own once the output looks right.

BANKING PIPELINES (synthetic, fake):
- Onboarding: retail checking, stages document capture, selfie/liveness, sanctions screen, manual KYC review for medium risk
- Transaction monitoring: wire and ACH rules, 1,200 alerts/month (fake), 68% closed as false positive
MANUAL REVIEW QUEUES (fake): KYC review queue avg 2.1 days, worst 9 days; alert disposition avg 5.4 days, worst 21 days; 6 analysts
BSA PROGRAM DOCS MAPPED (fake): CIP policy ref BSA-TEST-01; CDD refresh ref BSA-TEST-02; SAR workflow ref BSA-TEST-03
LATENCY BUDGET (fake): onboarding decision soft cap 300ms for edge checks; payment release hard cap 500ms

Pairs with on deepidv

Sources & further reading

FAQ

What did FinCEN's August 3, 2026 UBS enforcement action change for BSA programs?

FinCEN assessed a $125 million penalty against UBS for Bank Secrecy Act violations and framed the findings around outcomes: whether suspicious activity was actually intercepted, not whether policies existed on paper. Manual review queues that let flagged activity age for days were treated as program weaknesses in their own right. This prompt grades your pipelines against those criteria and designs the automated replacements.

What counts as a manual review bottleneck?

Any point where a flagged onboarding application or transaction waits on a human queue before a disposition: alert backlogs, dual-approval checks with no time bound, and escalation paths that cross teams. Luna inventories each one, measures its disposition time, and maps it to the enforcement criterion it puts at risk.

Are sub-150ms telemetry loops compatible with BSA record-keeping requirements?

Yes. Each automated verification loop the prompt configures emits a decision record with the signals checked, the thresholds applied, and the outcome, retained in an examiner-ready schema. Automation changes who executes the check, not what evidence is kept.

Can I use this prompt outside the deepidv dashboard?

Yes. The structure works in Claude, ChatGPT, or Gemini as a self-assessment framework and will return the criteria map, bottleneck inventory, and loop configuration. Live pipeline pulls, alert-queue metrics, and automated loop deployment only work when it runs inside the deepidv dashboard.

Run it with live verification data

These prompts work in any LLM. Inside the deepidv dashboard, Luna, Arbiter, and Arc run them against your real sessions, screening lists, and audit trails.

Book a Demo