Prompt Types

Text

Purpose: Extract or produce focused textual analysis up to one paragraph in length
Output: Structured text response with automated evidence links to supporting passages in the source document
Use cases: Summarise key findings, extract methodology descriptions, identify main arguments, describe compliance status
Workflow role: Terminal output only - cannot be used as conditions for other prompts

Tip: Use for extracting or generating up to a paragraph of text with evidence citations.

List of Values

Purpose: Extract discrete, short-form data points for categorisation and filtering each prompt targets one category.
Output: Structured list of brief items (titles, names, categories) with evidence links
Use cases: Extract company names, identify document types, list key stakeholders, catalogue mentioned regulations
Workflow role: Responses can serve as conditions for subsequent prompts and document filtering

Tip: Use to return individual datapoints from the document.

Range of Values

Purpose: Extract numerical data from documents or assign numerical scores based on specified criteria
Output: Numerical value only (no currency symbols or units), with evidence links supporting the extraction or assessment
Use cases: Extract wages, prices, quantities, percentages; rate document relevance, assess compliance levels, score quality metrics, evaluate risk factors
Workflow role: Responses can serve as conditions for subsequent prompts and populate filtering sliders for document sorting

Tip: Great for filtering a large dataset using numeric values.

Generative

Purpose: Produce comprehensive, unstructured analysis without length constraints
Output: Extended responses without evidence linking functionality
Use cases: Detailed summaries, comprehensive analysis, creative interpretation, extensive commentary
Workflow role: Terminal output only - cannot be used as conditions for other prompts

Tip: Instruct AI to provide document references with its response.

Yes/No

Purpose: Provide binary classification decisions
Output: Boolean response with evidence links supporting the determination
Use cases: Compliance verification, eligibility assessment, presence/absence determination, threshold validation
Workflow role: Responses can serve as conditions for subsequent prompts and document filtering

Tip: Separate documents or trigger follow on prompts based on y/n responses.

Workflow Integration

Playbook System: All prompts are pre-written and organised into playbooks, which serve as reusable analytical frameworks that users can apply to any document set.

Sequential Processing: Responses from Yes/No, Range, and List prompt types can be referenced as input conditions for subsequent prompts within the same playbook, enabling complex analytical workflows.

Document Filtering: Yes/No, Range, and List outputs automatically integrate with the platform's filtering system, allowing users to sort and narrow document sets based on extracted criteria.

Evidence Integration: All prompt types except Generative automatically generate evidence links to support assertions made in responses. This occurs in the UI after analysis is complete through the platform's built-in functionality, providing traceability to source material.

Common Questions

How do I handle unexpected or incorrect prompt responses?

During verification, you can edit AI responses for all prompt types. For List of values, you can remove incorrect items with the trash icon or add new categories. For Yes/No, you can toggle the correct answer. For Range of values, you can adjust the numerical value directly. Use cntrl-f in the document viewer to find supporting text and place it in the evidence field. 

If You're Still Stuck...

If you're unsure which prompt type to use:

1. Consider the format of the answer you need (yes/no, categories, numbers, text)
2. Determine if you'll need to filter documents based on the response
3. Create a test prompts with different prompt types to see which produces the most useful results

Still need help? Contact us via Zendesk.

Related Articles

• How do I create and use Flows?
• Understanding Playbooks 
• How do I set up conditional logic in Playbooks?
• How do I use the hash-linking system in Playbooks?