Prompts are the phrases used as input into generative AI models like Reggi, and crafting successful prompts will enable you to more efficiently leverage the platform.
Tips for Effective Prompts
Here are some tips for effective prompt engineering
Understand the capabilities and limitations of the model. Reggi can be used to identify and extract requirements, summarize laws and regulations, and answer general purpose compliance questions. Current limitations are include here.
Be as specific and detailed as possible: More detailed questions will lead to more detailed responses. If you do not receive the answer you were looking for, try adding more context and information to the question, like jurisdiction, industry, or use case.
Use a variety of prompt types. Reggi responds differently depending on the specific type of question. To see a list of prompt types as well as their examples, read below. Try varying your prompt types to improve results.
Test and refine prompts to improve their effectiveness. Try tweaking your questions by adding more information, different formats, and new phrasing to improve effectiveness. Save effect prompts by clicking the Save icon within the History section at the bottom right of the Reggi screen.
Supported Prompt Types and Examples
Summarization: Useful for summarizing a given authoritative document on Regology.
Example: Summarize 6.2 VACV Chapter 4, Article 2
Limitations: Summarizations at a high-level (Title, Provision, Etc.) may exclude authoritative documents depending on the number of authoritative documents.
Explanation: Useful for explaining a complex concept or what all of the steps for complying with a regulation would entail.
Example: Explain the process for reporting incidents to OSHA in CA
Extraction: Use this prompt type to extract key information like Requirements, Exemptions, Penalties, and more.
Requirements: What are the requirements for 26 CFR 1.11-1?
Penalties: What are the penalties of noncompliance with 26 CFR 1.11-1?
Exemptions: What are the exemptions of 26 CFR 1.11-1?
Ranking: Useful for prioritizing requirements.
Example: "Rank the sports betting requirements in Colorado in order of severity."
Scenario-Based: For outlining a specific scenario and identifying what steps need to be taken to be in compliance.
Example: If a company is advertising online for sports betting in Massachusetts, what precautions should they take?
Experimental Prompt Types and Examples
Note: These prompts are still works in progress but can be leveraged to generate answers.
Drafting Compliance Objects: For building out the first draft of a policy, control, or risk for your compliance program.
Example 1: Draft a policy for complying with CCPA.
Example 2: Draft a Risk for non-compliance with CCPA. Include a Title, Description, and Penalty Amount.
Comparison: For comparing requirements, penalties, or other items across multiple jurisdictions:
Example: "Compare and contrast the requirements for fantasy sports in Arizona and Colorado”
Limitations: When a topic is not supported in one jurisdiction.