There's numerous prompting techniques you can use to power Productised in the backend. In this guide we'll cover few-shot, zero-shot, chain-of-thought and prompt chaining methodologies.

The Value of Different Prompt Approaches

When building AI-powered products that generate dynamic outputs from form inputs (e.g., personalized roadmaps, checklists, plans, or insights), how you design your prompts matters.

Zero-Shot Prompting (Simple & Direct)

What it is:
You give a clear, specific instruction to the AI without showing any examples.

Use when:

• Your task is straightforward (e.g. summarizing, listing, rephrasing).

• You want to minimize complexity and keep outputs simple.

AI Product Use Case Example:
You’ve built a LinkedIn Bio Generator with one AI node. Your prompt might look like:

"Write a short, professional LinkedIn bio based on the user's job title: {form:job_title}, target audience: {form:audience}, and tone: {form:tone}. Keep under 3 sentences."

Best for:

• First drafts

• Simple content

• Rewording and formatting tasks

Few-Shot Prompting (Provide Examples)

What it is:
You give a few short examples of what you want before asking the model to create a new version based on new inputs.

Use when:

• You want the AI to mimic a specific tone, style, or format.

• You’ve found that outputs drift or vary too much.

AI Product Use Case Example:
You’ve created a Testimonial Writer product. Your prompt might include:

Example 1:
Client: “I loved how quickly things moved.”
Output: “Impressed by the fast turnaround and responsive service.”

Example 2:
Client: “They understood our vision.”
Output: “Their ability to grasp our vision from day one was remarkable.”

Now, write a testimonial for: {form:client_quote}

Best for:

• Mimicking tone/voice

• Domain-specific content (e.g. therapy, design, coaching)

• Templates with a specific structure

Tip: Keep examples short, and vary them slightly to show range.

Chain-of-Thought Prompting (Step-by-Step Reasoning)

What it is:
You guide the AI to break down the task into steps or logic, often by asking it to structure its thinking or rationale.

Use when:

• Your product requires evaluation, analysis, or comparison

• You want intermediate reasoning (e.g., priorities, trade-offs)

AI Product Use Case Example:
You’ve created a Goal Readiness Scorecard. In your prompt:

“Evaluate the user's goal clarity, motivation, and realism using their inputs:

• Goal: {form:goal}

• Clarity rating: {form:clarity}

• Motivation rating: {form:motivation}

• Realism rating: {form:realism}

Step 1: Identify potential strengths and challenges.
Step 2: Summarize in 2 sentences why the user may or may not be ready.
Step 3: Suggest 2 next steps.”

Best for:

• Planning tools

• Diagnostic assessments

• Strategic recommendations

Tip: Use numbered sections or subheadings in your AI output format to reflect the reasoning process.

Prompt Chaining (Complex Workflows)

What it is:
You divide a complex task into sequential steps. The output of one becomes the input of the next.

Use when:

• The final output requires structured transformation or multiple stages

• You want greater control, modularity, or reusable logic

AI Product Use Case Example:
You’re building a Personalized Business Plan Generator with a prompt that incorporates 3 sequential steps

Step 1 – Extract Core Strategy
Prompt uses: {form:goal}, {form:niche}, {form:audience}
Output variables:

• core_strategy

• positioning_summary

Step 2 – Develop Tactical Plan
Prompt uses: {core_strategy}, {positioning_summary}
Output variables:

• lead_gen_plan

• delivery_plan

• monetization_model

Step 3 – Output Final Plan
Prompt uses: {lead_gen_plan}, {delivery_plan}, {monetization_model}
Final Output: A 2-page, formatted business blueprint

Exemplar Completed Prompt With Chaining

You are a seasoned strategic planning assistant creating a Personalized Business Plan for a client.

The plan must be generated in three internally chained stages (Extract Core Strategy → Develop Tactical Plan → Output Final Plan) but returned as one cohesive 2‑page blueprint ready for display.

Tone: Clear, professional, actionable.
Audience: Entrepreneurs and small‑business owners.
Format: Bold section headings, short paragraphs (≤ 80 words), bullet points (≤ 16 words), and scannable layout.

✳️ Form Inputs

• Primary Goal: {form:goal}

• Niche/Industry: {form:niche}

• Target Audience: {form:audience}

✅ Chain the Logic Internally (no separate nodes)

Step 1 – Extract Core Strategy

• Restate the business goal and niche clearly.

• Identify the unique positioning and differentiators for the stated audience.

• Internally produce:

  • core_strategy = 1–2 sentence summary of the core strategy

  • positioning_summary = 3 bullet snapshot of positioning

Step 2 – Develop Tactical Plan
Using core_strategy and positioning_summary you just derived:

• Outline a lead‑generation approach tailored to the target audience.

• Outline a delivery/operations approach aligned to the goal and niche.

• Suggest a monetization model suited to the business type.

• Internally produce:

  • lead_gen_plan = 3 bullet lead generation plan

  • delivery_plan = 3 bullet delivery/operations plan

  • monetization_model = 3 bullet monetization model

Step 3 – Output Final Plan (Single Deliverable)
Combine all prior internal outputs into a cohesive 2‑page, formatted business blueprint with the following named output variables:

• page1_executive_overview = Executive Overview (uses core_strategy)

• page1_positioning_snapshot = Market Positioning Snapshot (uses positioning_summary)

• page2_lead_generation = Lead Generation Plan (uses lead_gen_plan)

• page2_delivery_plan = Delivery/Operations Plan (uses delivery_plan)

• page2_monetization_model = Monetization Model (uses monetization_model)

• page2_next_steps = 3 actionable next steps drawn from the full plan

Return all of the above variables as distinct, clearly labeled sections ready for rendering. Each section should include a bold heading followed by its content block.

Best for:

• Blueprints, playbooks, strategic frameworks

• Products with multiple pages or tabs

• Use cases where each step builds on the last

Tip: Always define your output variables clearly and test before chaining.

Which Prompting Style Should I Use?

✅ Prompt Design Best Practices

1. Always define the task. Avoid open-ended instructions like "analyze this." Say exactly what you want.

2. Specify the tone and audience. Example: “Write in a professional but warm tone for HR leaders.”

3. Set formatting expectations. Bullet points, section headers, word limits—all help reduce ambiguity.

4. Use placeholder variables consistently. Always use {form:...} or {node_output:...} in your prompts.

5. Avoid overload. Keep prompts lean—only include what's needed at each stage.

🔧 Testing Tips in AI Product Builders

• Start with zero-shot. It’s the simplest and often works surprisingly well.

• Add examples if output quality drops. Use few-shot only when necessary.

• Split big prompts. If your output is messy, break into a prompt chain.

• Use output variables. Don’t rely on raw text — structure your AI node outputs for better rendering.

💬 Common Issues & Fixes

🧩 Final Checklist

✅ Clear task and tone
✅ Specific format or output structure
✅ Proper use of input variables
✅ If multi-stage: prompt chaining with structured variables
✅ Tested across at least 3 input scenarios

By mastering these four prompting styles, you can design AI products that are not only dynamic and personalized — but also consistently accurate, valuable, and brand-aligned.