Document type

Jurisdiction

Target audience

Use LEAP fields to automatically populate prompt templates with content loaded in a LEAP matter, including LEAP Tables. Using a LEAP field is the first preference to populate prompt templates with important information, such as parties’ names, contact details and relevant dates. See further information about LEAP fields - LEAP fields - By Lawyers Manual - Confluence

Use square brackets for placeholders so the AI can clearly distinguish between static text in the template and the sections that must be populated with information from the LEAP matter.

Use clear, precise, and concise language throughout the legal document.

Avoid legal jargon or overly complex phrasing unless necessary for legal accuracy.

Define any technical or industry-specific terms clearly within the document.

Maintain a tone of neutrality and objectivity, focusing on facts rather than emotions or opinions. Do not include any assumptions or speculative statements.

Each paragraph should be confined to a distinct subject matter.

Each paragraph and subparagraph should have a line break in between them.

Each paragraph must be numbered sequentially in the document (e.g., 1, 2, 3, etc.).

Maintain paragraph numbering continuity throughout the document across all sections.

Headings are not to be numbered.

Ensure that the numbering sequence is consistent and does not restart within different sections or after headings.

Use clear, descriptive headings as per the template.

Do not number the headings.

Ensure the headings are formatted in bold.

The legal document should focus on presenting factual information rather than making legal arguments or conclusions unless specifically required (e.g., in legal submissions or briefs).

The document must not include any references to legislation, case law, or specific legal acts unless explicitly requested by the user.

Spelling: Australian English (e.g. behaviour, organised)

Date Format:

Spelling: British English (e.g. behaviour, organised)

Date Format:

Spelling: American English (e.g. behavior, organized)

Date Format:

Auto - For CA-V2 models, Auto is set to use the entire matter by default. For LA-V2 models where Fact Pages are included in the prompt, Auto reads and generates Fact Pages, then runs the prompt against those Fact Pages.

Entire matter – The prompt is run on the entire matter. This option is suitable for most prompts.

Folders – The prompt is run on one or more folders. This option is suitable for prompts where you want to only use the information contained within a specific folder/s. For example, a prompt summarising financial disclosure may be set to read only the financial disclosure folder/s.

Documents – The prompt is run on one or more documents. This option is suitable where you want to summarise and/or compare two or more documents, such as comparing offers or inconsistencies in affidavits.

Fact Pages – The prompt is run on one or more Fact Pages (factual issues in the matter). There is also the option to run the prompt on Fact Pages with specific labelled facts.

You should consider including Preliminary Questions in the prompt to guide the user on whether it should be run on folders, documents, or Fact Pages.

If a user selects a folder, document, or Fact Page, the output will be generated solely from the content contained within those selected items.

Prompt performance can vary between models. Each time the model changes, some level of re-tuning of the prompt may be required.

Avoid using bold or ALL CAPS for emphasis in the prompt. These formatting styles are not necessary in newer models and may result in the model over-prioritising those instructions (e.g. IMPORTANT RULES may prioritise those rules to the exclusion of other rules).

Be explicit about your desired output formatting. For example, clarify if you want numbered lists, bullet points, paragraph structure, headings in bold, or tables. However, note that AI cannot control font, line spacing, or precise page layout - these should be managed at the document formatting stage (e.g. via LEAP's Insert Document function).

Use detailed placeholder instructions. Specify what to include and what to exclude. For example:

Note that the internal prompt contains information about accuracy, avoiding hallucinations and final review requirements. This does not need to be included in the legal prompts.

Clearly instruct the AI on how it should behave (e.g. write in formal legalese, summarise only, use plain English). Check for ignored or misinterpreted instructions and refine as needed.

Test and refine prompts with multiple sample inputs before going live.

Verify the output when testing AI Prompts.

Resources:

Provide feedback or log a case (bugs) on prompts to CORTO through the internal portal – https://provider.docorto.ai not via email/slack. Provide as much information as possible about the issue, with debug information, screenshots and screen recordings.