AI for Compliance Officers

You are a compliance writer. Rewrite the policy below in plain American English at an eighth-grade reading level for [AUDIENCE, E.G. ALL EMPLOYEES]. Preserve every requirement, prohibition, exception, and approval step.

Output: 1) the rewritten policy with sentence-case headings, 2) a table mapping each requirement in the rewrite to the clause of the original it came from, 3) a list of anything in the original you could not place, 4) five scenario-based knowledge-check questions with answers. Do not add requirements that are not in the original.

Original policy:
[PASTE POLICY]

Tip: Item 3 is where dropped exceptions show up; read it first.

You are a regulatory analyst. Answer only from the uploaded rule text. Our business: [DESCRIBE INDUSTRY, SIZE, PRODUCTS, JURISDICTIONS, AND LICENSES].

Provide, with section citations and page numbers: 1) which provisions apply to us and why, 2) which do not and why, 3) every compliance or effective date, 4) definitions that determine applicability, 5) recordkeeping and reporting obligations, 6) open questions the text does not resolve. If the document is silent, say NOT ADDRESSED. Do not draw on outside knowledge or prior versions of the rule.

Tip: Run it in NotebookLM or a Claude Project with the final rule uploaded; then verify the dates against the Federal Register entry.

You are an instructional designer for a corporate compliance team. Topic: [TOPIC, E.G. ANTI-BRIBERY FOR SALES TEAMS]. Audience: [AUDIENCE AND WHAT THEY ALREADY KNOW]. Policy excerpt: [PASTE THE RELEVANT RULES]. The three situations that actually happen here: [DESCRIBE THEM].

Produce a 20-minute session outline, three scenario scripts that each end in a decision with the correct answer and the policy basis, a five-question knowledge check, and a one-page takeaway sheet. Plain language, no legal citations in the learner-facing text.

Tip: Check every scenario resolution against the policy yourself before it goes to the LMS.

You are an investigations analyst working under counsel. From the de-identified notes below, produce: 1) a dated chronology of events with the source note for each entry, 2) contradictions between sources, 3) gaps where evidence is missing, 4) follow-up questions by witness, 5) documents to request.

Do not assess credibility, assign fault, or conclude anything; state only what the notes say and do not say.

Notes:
[PASTE DE-IDENTIFIED INTERVIEW AND DOCUMENT NOTES]

Tip: Use only in a tool your legal team has approved for privileged work, and keep the output in the case file under the same controls.

You are a compliance analyst reviewing a request to use an AI tool. Request: [TOOL NAME, VENDOR, INTENDED USE, TEAM, DATA TYPES INVOLVED].

Produce: 1) the data classification questions to answer before approval, 2) vendor questions on input retention, training on customer data, subprocessors, deletion, and security attestations, 3) laws or regulations to check given the data types and use (list them for me to verify; do not summarize them), 4) the risk rating criteria I should apply, 5) conditions of use to attach to an approval. Do not approve or reject; structure the review.

Tip: Paste the vendor's actual terms into a second turn and ask which of the questions in item 2 they answer.