The Iterative Prompting Loop: You're Not Done After One Try

Most recruiters think of prompting as one-shot: ask, get answer, move on. But prompting is iterative. You ask, review, refine, ask again. Each cycle gets you closer to what you actually need.

This lesson teaches you the feedback loop. You'll learn how to conduct multi-turn conversations with AI, when to iterate vs. start over, and how to move from "this is okay" to "this is exactly what I need."

The Iterative Loop: 7 Steps

Step 1: Ask (Your Initial Prompt) - Based on context, task, constraints, specificity

Step 2: Get Output - AI produces something

Step 3: Review (5 Questions)
- Accurate? (Did AI hallucinate?)
- Complete? (Did AI answer fully?)
- Relevant? (Specific to my situation?)
- Tone/voice? (Sounds like me?)
- Actionable? (Can I use this?)

Step 4: Assess (Decision Point)
- Worth refining? (70%+ there) → Go to Step 5
- Start over? (Fundamentally wrong) → New prompt

Step 5: Identify the Gap
What's wrong or missing exactly?
- Content gap (missing detail)
- Tone gap (too formal/casual)
- Structure gap (wrong format)
- Accuracy gap (got facts wrong)

Step 6: Refine Your Prompt
Add information about the gap. Don't rewrite everything. One or two sentence additions usually fix it.

Step 7: Review Again - Is it closer? Still missing something? Go to Step 6 or accept it.

Multi-Turn Conversation Techniques

Technique 1: Context Building

Early prompts establish context. Later prompts build on that context.

Example Multi-Turn:
Turn 1: "You're a recruiter at [Company]. [Context about company]."
Turn 2: (referencing earlier context) "For this candidate, what would a good outreach message look like?"
AI understands context from Turn 1, applies it to Turn 2

Technique 2: Incremental Refinement

Turn 1: "Draft an interview question about time management."
Turn 2: "Good, but make it more challenging. Add a follow-up probe."
Turn 3: "Perfect. Now generate 4 more similar questions."
Each turn builds on previous, no need to re-explain basics.

Technique 3: Example-Based Refinement

Turn 1: "Draft an outreach message."
AI Output: Generic result
Turn 2: "Here's an example of what I mean [paste good message]. Try again using this style."
AI now has concrete example, produces better output.

Technique 4: Constraint Tightening

Turn 1: "Generate interview questions."
AI Output: Generic questions
Turn 2: "Good start. These are too broad. Focus on specific technical skills. Probe for depth, not breadth."
AI tightens constraints, produces more specific output.

Iteration Decision Framework

Common Iteration Patterns and Fixes

Pattern 1: Too Generic

First output: "She was great. Really smart. Definitely hire her."
Problem: Could apply to anyone
Refinement: "Make this more specific. Reference her actual background from these notes: [paste notes]"
Second output: Much more specific

Pattern 2: Too Technical

First output: "Implement comprehensive automated decision-making framework..."
Problem: Too complex for the audience
Refinement: "Simplify this. I'm explaining to non-technical hiring managers. Use plain language."
Second output: Simpler, clearer

Pattern 3: Missing Key Detail

First output: Message about candidate, but doesn't mention their specific project work
Problem: You wanted personalization on that project
Refinement: "Good start. Make sure to reference their work on [specific project] - that's what grabbed our attention."
Second output: Now mentions project

Pattern 4: Wrong Format

First output: Narrative paragraphs
Problem: You needed bullets
Refinement: "Same content but format as bullets instead of paragraphs"
Second output: Bulleted version

When to Stop Iterating

At that point, take the AI output and manually edit the last 15%. That's faster than infinite iteration.

Multi-Turn Conversation Examples

Example 1: Building Screening Rubric

Turn 1: "I'm screening for a Senior PM role. Key requirements are [list them]. Create a scoring rubric."
Turn 2: "Good. Now make it more specific. How exactly would I assess each criterion?"
Turn 3: "Perfect. Now give me example answers that would score 3/3 vs 1/3 on [specific criterion]."
By Turn 3, you have detailed, practical rubric without starting from scratch.

Example 2: Iterating Interview Questions

Turn 1: "Generate 5 behavioral questions for assessing technical depth in [role]."
Turn 2: "Good. Q3 is too broad. Make it more specific. Include follow-up probe."
Turn 3: "Got it. Now create a scoring guide: what's strong response vs weak response?"
Each turn builds on previous. Questions improve. You add scoring guide without restating role.

Documentation: Track What Works

Keep a "Prompt Journal":

• What prompt worked well? (Save it)
• What iteration fixed the problem? (Document the refinement)
• What constraint always helps? (Note it)
• What phrasing produces best output? (Remember it)

After 20-30 prompting sessions, patterns emerge. You'll notice: "Adding [X constraint] always improves output" or "When I say [Y phrase], AI understands better."

This becomes your prompting playbook.

FAQ: Iteration and Refinement

How many times should I iterate on a single prompt?

Typically 2-3 times. More than 3, you're spending time iterating that could go to manual editing. If you're iterating more than 3 times, consider starting over with better initial prompt.

Should I ask AI to "try again" or refine my prompt?

Always refine your prompt. "Try again" without changing prompt usually produces similar output. Refining means identifying what was wrong and telling AI specifically how to fix it. That produces different, better output.

Can I reuse context across multiple prompts in a conversation?

Absolutely. Set context once ("You're a recruiter at..."), then ask follow-up questions. AI remembers context, so you don't repeat it. This is the power of multi-turn conversations.