> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bitscale.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Output Formatting

> Output formatting allows you to - Enforce structure on AI responses - Build reliable, multi-step workflows - Prevent hallucination-driven failures - Safely automate downstream actions

<iframe width="100%" style={{aspectRatio: "16/9"}} src="https://www.youtube.com/embed/VdYoEvdnzOA?si=U_YibEP1x0sP7Zr8" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen />

When building automated workflows, **structure matters**.

Any response you generate, especially from AI models, needs to follow a predictable format so it can be:

* Referenced by later steps
* Used in conditions or classifications
* Pushed reliably to CRMs, grids, or integrations

AI models are powerful, but by default they can:

* Change response formats
* Add extra text
* Reorder information
* Hallucinate structure even with strict prompts

To solve this, Bitscale provides **Output Formatting**, a built-in way to enforce structured AI responses across all AI enrichments.

***

## Why Output Formatting Is Important

Without structured outputs:

* Workflows break unpredictably
* Downstream enrichments fail
* CRM pushes become unreliable
* Automation becomes fragile

With output formatting:

* Every run produces the same schema
* Columns are always consistent
* AI results can safely drive logic and actions

***

## Where Output Formatting Is Available

Output formatting is available on:

* **Bit Agent**
* **GPT / Claude / AI enrichments**
* **Any text-based AI step**

You’ll find it under **Set Output Format** in the enrichment configuration.

***

## Fields vs Lists

Bitscale supports two output types:

### Fields

* Used for **single values** or **comma-separated values**
* Best for:
  * Industry
  * ICP classification
  * Boolean flags
  * Short summaries

**Example:**

```
Industry: SaaS, DevTools
Target Demography: Mid-market B2B
```

***

### Lists

* Used for **multiple structured entries**
* Each item is treated as a separate entity
* Best for:
  * Event attendees
  * Customers
  * Technologies
  * People lists
  * Anything you may want to explode into rows later

**Example:**

```
Customer 1: Name, Company
Customer 2: Name, Company
```

If you plan to export or explode results into another grid, always use **Lists**.

***

## Manually Defining Output Formats

You can manually define:

* Field name
* Data type (text, boolean, number, etc.)
* Output structure

This gives maximum control but requires clarity on what the model should return.

***

## Using “Generate with AI” (Recommended)

For most users, the easiest and safest approach is **Use AI**.

### How it works

* Bitscale reads your prompt
* Understands the intent
* Automatically generates the correct output schema

### Example

Prompt:

> Go through @company\_website and find their target demography and customers

Click **Use AI**, and Bitscale generates:

* Target Demography
* Industry
* Customers

All correctly typed and structured.

This dramatically reduces prompt engineering effort and prevents format drift.

***

## Running and Storing Structured Outputs

Once the enrichment runs:

1. The AI response is forced into the defined structure
2. Each output field becomes a reusable column
3. These columns can now be:
   * Referenced in later prompts
   * Used in run conditions
   * Exported to CRMs
   * Passed to other grids or integrations

No matter how many times the enrichment runs, the structure remains consistent.

***

## Best Practices

* Always use output formatting for **AI-driven workflows**
* Use **Fields** unless you explicitly need row-level expansion
* Prefer **Use AI** unless you have very strict schema requirements
* Treat structured outputs as **foundational workflow primitives**

***