AI Prompt Writing Hints & Tips
Writing effective prompts is key to unlocking the full potential of AI-powered workflows in Intellectible. Whether you're summarizing documents, extracting structured insights, or guiding proposal writing, a well-crafted prompt ensures clarity, consistency, and relevance.
This guide introduces the CLEAR Framework, advanced strategies, key configuration settings, and real-world prompt templates—so you can build smarter automations from day one.
AI Settings Inside Intellectible
When configuring an AI Write or AI Write Data node, you can adjust key settings to influence tone, output length, and consistency. These settings are especially important when using multiple AI nodes together in complex workflows.
Temperature
Temperature controls the creativity and variability of the AI’s output.
| Temperature Value | Effect | Use Case |
|---|---|---|
| 0.0 – 0.3 (Low) | Very focused, deterministic, and factual. | Ideal for structured tasks like document summarization, data extraction, or matching requirements. |
| 0.5 – 0.7 (Medium) | Balanced and natural; adds a touch of variation without losing structure. | Great for general-purpose writing, short summaries, or workflows that require a human-like tone. |
| 0.8 – 1.0+ (High) | Creative, open-ended, less predictable. | Best for brainstorming, idea generation, or crafting varied responses. Not ideal for production workflows. |
Max Tokens
Max Tokens sets the maximum length of the AI’s output. Set token limits based on expected output, but allow room for growth. If you notice responses being cut off, increase the token limit accordingly.
(1 token ≈ ¾ of a word. For example, 100 tokens ≈ 75 words)
| Token Count | Approximate Output | Use Case |
|---|---|---|
| 500–1000 | Short summaries, bullet points, answers | Compact document overviews or short form responses |
| 1000–2000 | Medium-length structured content | Grant section summaries, multi-part responses |
| 3000–6000+ | Full document breakdowns or long-form outputs | Deep analysis, full response generation, multiple AI nodes working together |
Seed
Seed determines the AI’s internal randomness. In Intellectible, the Seed setting accepts a numeric value to control randomness. A fixed seed ensures the same prompt consistently produces the same output. A high/random seed allows variation across runs - especially in production workflows where reliability matters.
| Seed Value | Value Range | Behavior | Use Case |
|---|---|---|---|
| Low (Fixed) | 1-999 | Repeatable, consistent output | Use when multiple AI nodes are linked or when outputs are reused in logic |
| High (Variable) | 1000+ or Randomized | Allows variability across runs | Useful for creativity, generating varied ideas, or exploration tasks |
In Intellectible, workflows often involve multiple AI nodes working together across complex processes. To ensure consistency and reliable results — especially when outputs are reused or connected it's best to keep Temperature low and use a fixed Seed. Before increasing creativity settings, consider how small variations can compound across workflows.
Recommended Defaults for Most Workflows
| Setting | Recommended |
|---|---|
| Temperature | 0.5–0.7 – Balanced and reliable |
| Seed | Fixed value like 1, for consistent output |
| Max Tokens | 1000–2000 for summaries, 5000+ for longer or multi-step workflows |
The CLEAR Framework for Better Prompts
Use the CLEAR framework as a foundation for every prompt you write inside AI Write or AI Write Data nodes:
| Principle | What It Means | Example |
|---|---|---|
| C – Concise | Eliminate unnecessary words. Focus on the core request. | Initial: “Can you explain the different ways our proposal might be improved for submission?” Refined: “Suggestions to improve this grant proposal before submission.” |
| L – Logical | Structure the request clearly to guide the AI’s flow. | Initial: “How do we hire people?” Refined: “List 3 steps to improve our technical recruitment process, with a short explanation for each.” |
| E – Explicit | Be specific about what you want and what to avoid. | Initial: “Summarize this company profile.” Refined: “Summarize this company profile in 5 bullet points. Exclude financials and awards.” |
| A – Adaptive | Modify your prompt based on the AI’s previous responses. | Initial: “Does the company align with the funding call?” Refined: “List each solicitation requirement and explain how the company meets or doesn’t meet it.” |
| R – Reflective | Review the output, then refine the prompt to get more targeted insights. | Initial: “Summarize the uploaded market research report.” Refined: “Summarize the key findings of the market research, focusing only on competitor analysis and industry trends.” |
Advanced Prompting Strategies in Intellectible
Use these strategies to improve the quality, consistency, and usefulness of AI outputs:
-
Be Detailed When It Matters
Add context like the AI’s role, desired format, and exclusions.
Example: “You are a grant reviewer. Summarize the company’s capabilities in bullet points. Exclude background info.” -
Guide the AI’s Reasoning
Ask it to explain matches or follow a step-by-step structure.
Example: “For each requirement, explain which company capability matches and why.” -
Request Multiple Options
Encourage broader coverage by asking for variations or multiple outputs.
Example: “List three key risks and suggest a mitigation strategy for each.” -
Use Negative Instructions
Prevent irrelevant output by stating what not to include.
Example: “Summarize the proposal’s objectives. Do not include historical context or general background.”
Prompt Templates You Can Use
Jumpstart your workflow creation with these tested examples:
Document Summarization
“Summarize this document in 3 bullet points per section, highlighting key eligibility criteria, funding amounts, and submission deadlines.”
Capability Matching
“Match the company’s capabilities to each solicitation requirement. Use a 2-column table: Requirement | Matching Capability.”
Risk Identification
“List potential technical or budgetary risks mentioned in this document. For each, suggest a mitigation strategy.”
Form Output Generation
“Using the form inputs, generate a one-paragraph project summary that incorporates all entered values. Write clearly and professionally.”
Final Tips
- Start broad, then refine. Tweak prompts to improve structure and clarity.
- Use formatting examples inside prompts to control how results are presented.
- Leverage dynamic variables like
{{form_input}}to personalize prompts across workflows. - Test multiple variations and compare results to determine what works best.