Skip to main content

​
What It Does

  • Custom Prompting: Generate AI responses based on any prompt and dataset—personalized row by row.
  • Structured or Free-Form: Choose between natural text or JSON-style structured output.
  • Supports Web Search: Enable external search for richer answers on supported models.
  • Column Auto-Mapping: Dynamically maps fields and avoids name collisions for smooth output.
  • Model Flexibility: Run on multiple OpenAI models including GPT-4o and GPT-4o-mini.

​
🏁 Getting Started

Ask AI node configuration screenshot
1

Add the Node

Drag & drop the Ask AI node into your workflow.
2

Write the Prompt

Use the Insert Input button or @column_name to reference values from your input.
3

Select Model

Choose from supported models like gpt-4o-mini (default) or gpt-4o.
4

Choose Output Type

Select Free Text (default) or Structured Output depending on your needs.
5

(Optional) Configure JSON Schema

For structured output, define expected fields using a simple JSON schema.
6

Run the Node

Click Run to generate AI responses for each row in your input dataset.

​
Inputs

​
🛠️ Required Fields

  • Prompt (✅)
    Write your instruction using @column_name references or Insert Input.
    Why it matters: The AI uses this row-by-row to tailor each result.
  • Model (✅)
    Choose between gpt-4o-mini (default), gpt-4o, or any other supported OpenAI model.
    Why it matters: Affects quality, speed, and credit cost.
  • Response Type (✅)
    Choose free_text (default) or structured_output.
    Why it matters: Controls whether you get one column (text) or multiple (JSON→columns).
  • Web Search Enabled (✅)
    Toggle On/Off for real-time context injection on supported models.
    Why it matters: Improves freshness of answers but adds cost + latency.

​
🎯 Optional Fields

  • Response JSON (⚪️)
    Required if structured_output is selected. Use sample JSON to define column layout.
  • Output Column Name (⚪️)
    Applies only to free-text output. Default is RESPONSE.
We also support Latest OpenAI model- gpt-5.1!!

​
Output

  • Free Text Output
    • Adds one column: RESPONSE (or renamed output)
  • Structured Output
    • Adds one column for each key in your JSON schema

​
How It Works

  1. Injects column values dynamically from each row into your prompt
  2. Sends request to your selected model (with web search if enabled)
  3. Parses output as free text or structured JSON
  4. Adds result(s) to each row of your table
  5. Automatically renames fields if name conflicts occur

​
🚀 Example Use Cases & Prompts

Use CasePrompt
Product Suggestions”Suggest products for @customer_name based on their history: @purchase_history”
Feedback Sentiment”Analyze sentiment for @customer_name based on: @feedback”
Lead Scoring Summary”Summarize activity for @lead_name and recommend next step.”
Job Title Classifier”Classify job title into Sales, Marketing, or Engineering: @title”
Category Mapping”Assign a product category to this item: @product_name”

​
✨ Pro Tips

Use Insert Input to pull in dynamic row data — it’s safer than typing and avoids name mismatches.
Run a sample first. Validate your prompt on 5–10 rows before scaling to hundreds.
Use structured_output if you want reliable, named columns you can filter, score, or export later.

​
⚠️ Important Considerations

Missing or incorrect column references = blank output. Use @column_name or Insert Input to avoid issues.
Web Search may increase latency and credit use — enable only when external context is truly needed.
If structured_output is selected, the Response JSON schema is required — it defines what columns to return.
Field collisions are resolved automatically using suffixes like _1, _2, etc.
Not all models offer websearch and structured output. Check model capabilities before configuring.

​
🛠 Troubleshooting & Gotchas

SymptomLikely CauseQuick Fix
No output generatedMissing column in promptUse Insert Input or double-check @references
All rows = same answerStatic prompt with no dynamic dataAdd @column_name to personalize per row
Slower than expectedWeb search enabledTurn off unless needed
Output columns missingMissing or malformed JSON schemaAdd a valid JSON schema

​
📝 FAQ

Yes — chain as many as you’d like for layered classification, insights, or scoring.
Not in the same node. Use separate nodes if you need both types.
The row is preserved, but the output column will show an error message.

​
💰 Pricing

Ask AI is billed per processed row:
ModelCredits / RowNotes
GPT-5.15Most advanced, top reasoning
GPT-55Near-frontier intelligence
GPT-5 mini3Efficient, strong reasoning
GPT-5 nano2Ultra-light, fast responses
GPT-4.13Balanced performance
GPT-4.1 mini2Lightweight, optimized
GPT-4.1 nano1Very fast, minimal compute
GPT-o34Legacy model, stable output
GPT-o4 mini4Improved mini-tier model
gpt-4o-mini1Fast, lightweight
gpt-4o4Premium model, deeper reasoning
Each Web Search toggle adds +2 credits/row to your base cost.

Add Ask AI to your workflow and turn raw data into answers, summaries, and classifications — all without switching tools. ⚡🧠