How to Convert Free Text Into Structured JSON in Batch in Excel

Why JSON Conversion Works Well in BatchGPT

The task is repetitive and schema-based: the same keys should appear for every row, even when the text varies. That makes JSON conversion a practical fit for an Excel add-in that runs the same prompt logic across many cells.

How to Use the BatchGPT Excel Add-in for This Workflow

1. Write the prompt that tells the add-in what to do with each selected cell value.
2. Select the Excel cells or range you want to process. For larger datasets, work in clean batches of rows.
3. Choose the output column and adjust optional settings such as reasoning effort or web search when the task really needs them.
4. Click Generate so the add-in processes each selected cell separately and writes the result to the output column you chose.
5. Review the results in Excel, refine the prompt if needed, and rerun only the rows that need another pass.

Prompt Example for Free-Text to JSON Conversion

The schema matters more than long instructions. Be explicit about keys, null handling, and formatting.

Convert this free text into valid JSON:


Required schema:
{
  "customer_name": string | null,
  "issue_type": string | null,
  "urgency": "low" | "medium" | "high" | null,
  "requested_action": string | null,
  "due_date": string | null
}

Rules:
- Return valid JSON only
- Use null for missing fields
- Do not invent values

Sample input row:

A2: Customer Anna Brooks reported an invoice mismatch and requested correction before March 10. Mark as high priority.

Sample output row:

{
  "customer_name": "Anna Brooks",
  "issue_type": "invoice mismatch",
  "urgency": "high",
  "requested_action": "invoice correction",
  "due_date": "March 10"
}

How to Keep JSON Output Predictable Across Rows

Consistency matters more than cleverness when the output is headed to another workflow.

• Lock down the exact keys and allowed enum values before you process a larger range of text rows.
• Use null for missing fields instead of optional free-form placeholders like unknown or maybe.
• Run a small batch first and validate the JSON format before scaling to a larger selection.
• If one source column contains very different text types, split them into separate schema-specific runs.

FAQ

Can I enforce a strict JSON schema for every row?

Yes. That is one of the main advantages of this workflow, as long as the schema is clearly defined in the prompt.

What happens when data is missing?

Use null rules so the same keys still appear in every response and downstream parsing stays predictable.

Can I use this for ticket notes or CRM logs?

Yes. Any free-text field that needs structured extraction is a good fit for this pattern.

Should I publish the JSON without checking it?

You should still validate the format and the extracted values on a sample set before you rely on it in downstream systems.