Autocode with Protobi AI

Protobi AI Autocode is a feature that automates the categorization of open-end survey responses. Instead of manually reading through hundreds of responses and building a codeframe by hand, Protobi AI Autocode analyzes your responses, identifies recurring themes, and applies structured codes — significantly reducing the time and effort required for open-end analysis.

This guide is for researchers and analysts who want to:

• Build a new codeframe from scratch
• Complete a partially coded dataset
• Apply an existing codeframe to new responses
• Generate clear labels and descriptions for existing codes
• Validate the quality of coding assignments before final delivery

Protobi AI Autocode supports seven operations — Create, Augment, Apply, Describe, Evaluate, Validate, and Clear — each designed for a specific stage of the coding workflow. The sections below explain what each option does and walk you through how to use them step by step.

The following table provides a quick overview of all available AI Autocode options:

Option	What It Does	Best Used When
Create	Generates a brand-new codeframe from scratch and tags all responses	No prior coding exists — starting fresh
Augment	Completes a partial codeframe without modifying existing codes	Some responses already coded — need to finish the rest
Apply	Applies your existing codeframe strictly — no new categories generated	Predefined codeframe must be followed exactly
Describe	Generates clear labels and descriptions for existing coded categories	Codes exist but lack clear definitions or documentation
Evaluate	Generates an independent AI reference solution and compares it against your current coding	After completing coding — QC validation before final delivery
Validate	Reviews individual assignments against the coding scheme to populate the QC column	Spot-checking individual assignments for quality
Clear	Removes all existing codes and assignments — returns responses to untagged state	Starting over completely — irreversible action

AI Autocode options & what they do

Create: Create and apply a new categorization given text responses

This option generates a brand-new codeframe from scratch based on your raw open-end responses. The AI analyzes patterns, themes, and keywords in the data to suggest logical categories.

This is best for questions where all responses are still raw open-ends and you have not implemented any coding yet, or for questions where you want to remove all existing codes and start from scratch using the AI's categorizations.

Augment: Augment and apply an existing partial categorization

This option finishes a partially coded dataset by applying AI-suggested categories to uncategorized responses only. It ensures consistency with your existing categories.

This option is particularly valuable when you have partially coded raw open-ended responses and require an efficient way to categorize the remaining data. By manually coding a subset of responses, you establish a clear codeframe, which guides the AI in applying consistent codes to uncategorized responses.

Apply: Apply (but not change) existing categories to given responses

This option strictly uses predefined categories to classify responses without modifying or generating any new codes. Ideal when you only need automated tagging.

This functions similarly to Augment, but with one key difference: Apply won't generate new categories. Use Augment instead if you already have some predefined categories but also want the AI to suggest additional ones while tagging unclassified responses.

Describe: Infer category labels and descriptions from an existing categorization

This option analyzes your existing categories and generates clear labels and descriptions for clarity. Useful for assigning definitions to coded categories for documentation purposes and clarity.

Note: The Describe function runs automatically when using Create, but does not activate when using Augment or Apply operations.

Evaluate: Appraise the quality of assignments

This option independently reapplies your existing coding scheme to generate a reference solution, then compares it against your current assignments — highlighting where the two agree and where they differ. Use this after completing your codeframe to validate coding quality before final delivery.

Validate: Manually spot-check coding assignments

Once you have run any AI Autocode option — Create, Augment, Apply — or completed manual coding, Validate allows you to review each assignment individually and confirm whether it is correct. User reads each response and its assigned code, then marks it using the QC column.

• Yes — Assignment is correct. No change needed.
• No — Assignment is incorrect. Unapply or reassign the response.
• TBD — Needs further review. Return to it before finalizing.

Clear: Clear all categories and assignments

This option removes all existing codes and assignments, returning responses to their original untagged state. Warning: This action cannot be undone. Clone your element before using Clear to preserve a backup of your existing coding.

Step-by-step guide to use Protobi AI Autocode

The following steps walk you through coding open-end responses in Recode (Advanced) using Protobi AI Autocode. Complete Steps 1–3 first, then follow the workflow for your chosen Autocode option.

Before you begin: You must have edit access to the project to use Recode (Advanced) and AI Autocode. Contact your project administrator if you do not have the required access level.

1. Getting started

Select the element where you want to apply Protobi AI Autocode.

2. Open Recode (advanced)

Once you've selected the open-end element, choose the Recode (advanced) button in the toolbar.

3. Choose your AI Autocode option

Press the "Autocode with Protobi AI..." button in the lower-right corner. The Autocode dialog will appear, where you can select the type of Autocode you want to apply.

Create a new codeframe

1. Start the process

   • Select the Create option and press OK in the dialog box.

2. AI analysis & coding

   • The system will automatically:
     • Analyze your responses.
     • Generate a customized codeframe with relevant categories.
     • Tag responses with the appropriate codes.

3. Review the results

   • Examine the suggested codes and their descriptions for accuracy.

4. Make adjustments (if needed)

   • Edit AI-generated codes and descriptions to match your terminology.
   • Unapply removes code tags from responses while keeping the code in your codeframe. Delete permanently removes the code and all its assignments from the codeframe.
   • Create codes manually and reassign responses to more accurate codes as necessary.

Tip: Use Create when exploring new data without a predefined codeframe. Also, before you begin clone your element to preserve the original raw responses. Save your project before applying AI Autocode.

Augment a partial coding

The Augment workflow follows the same steps as Create. The key difference is the AI will only add to your existing codes — not replace them.

1. Start the process

   • Select the Augment option and press OK in the dialog box.

2. AI analysis & completion

   • The system will automatically:
     • Analyze your partially coded responses.
     • Suggest additional categories to complete your codeframe.
     • Tag uncategorized responses while preserving your existing codes.

3. Review the results

   • Examine both the new and existing codes for consistency.

4. Make adjustments (if needed)

   • Edit AI-generated codes and descriptions to match your terminology.
   • Unapply removes code tags from responses while keeping the code in your codeframe. Delete permanently removes the code and all its assignments from the codeframe.
   • Create codes manually and reassign responses more accurate codes as necessary.

Tip: Use Augment when you have partially coded responses and want the AI to complete the remaining uncoded entries without modifying your existing codes.

Apply existing codes

1. Start the process

   • Select the Apply option and press OK in the dialog box.

2. Automated coding

   • The system will:
     • Strictly use your current codeframe.
     • Tag all responses without modifying or adding categories.

3. Review the results

   • Review the applied codes to confirm they have been assigned correctly.

4. Make adjustments (if needed)

   • Reassign responses to more accurate codes as necessary.
   • Responses that cannot be matched to any existing code will remain uncoded. Review these manually after the Apply process is complete.

Tip: Ideal for projects requiring strict adherence to an established codeframe. And Describe does not run automatically with Apply. Run Describe manually if needed.

Describe existing codes

1. Start the process

   • Select the Describe option and press OK in the dialog box.

2. AI analysis & description

   • The system will:
     • Analyze patterns in your coded categories.
     • Generate clear labels and definitions for each coded category.

3. Review the results

   • Review the suggested descriptions for accuracy.

4. Make adjustments (if needed)

   • Edit code descriptions as necessary to meet your expectations.

Tip: Use Describe to generate clear code definitions before sharing your codeframe with clients, stakeholders, or team members. It runs automatically when using Create. Run it manually when you need to generate or update descriptions for an existing codeframe.

Evaluate existing assignment

1. Start the process

   • Select the Evaluate option and press OK in the dialog box.

2. AI analysis & comparison

   • The system will:
     • Re-applies the existing coding scheme to generate an independent reference solution
     • Compare two coding solutions: the user's existing assignments against the AI's independently re-applied assignments.

3. Review assignment using the comparison toggle indicators:

   • In both schemes — Your coding and the AI reference agree on this assignment. High confidence; no action needed.
   • In current scheme only — This assignment exists in your coding but not in the AI reference. Review and decide whether to keep or remove it.
   • In reference scheme only — This assignment exists in the AI reference but not in your coding. Review and decide whether to add it.

4. Record your QC decisions

   Use the QC column to record your decision for each flagged assignment:

   • Yes — Assignment is correct. No change needed.
   • No — Assignment is incorrect. Unapply or reassign the response.
   • TBD — Needs further review. Flag for manual review before finalizing.

5. Make adjustments and save

   • Use the action buttons in the comparison panel to accept or reject each flagged assignment.
   • Reassign responses marked No to more accurate codes.
   • Review all TBD assignments manually before finalizing.
   • Press Save in the project toolbar to make all accepted changes permanent.

Tip: Use Evaluate after completing your codeframe to validate coding quality before delivering final results to a client.

Read the full guide on Evaluate Coding Quality →

Validate existing assignments

Note: Complete your coding first — using any Autocode option or manually — then use Validate to spot-check assignments.

1. Review each assignment

   • Go through each response and its assigned code in Recode (Advanced).
   • Ask: does this code accurately represent what the respondent said?

2. Record your QC decisions

   • Yes — Assignment is correct. No change needed.
   • No — Assignment is incorrect. Unapply or reassign the response.
   • TBD — Needs further review. Flag and return before finalizing.

3. Make adjustments and save

• Unapply or reassign any No assignments.
• Review all TBD assignments before finalizing.
• Press Save in the project toolbar to make all changes permanent.

Read the full guide on Evaluate and Validate Coding Quality →

Recode (advanced) user interface:

1. Status indicator

A progress meter and success notification appear in the upper right when Autocode is applied.

2. Code set panel

The panel displays all categories in your codeframe.
You can drag and drop codes from this panel to tag responses.

3. Code management

Edit code names, labels, and detailed descriptions.
Unapply removes tags from all responses while keeping the code in your codeframe. Delete permanently removes the code and all its assignments.

4. View code descriptions

Hover over codes to view their detailed descriptions (generated by the Describe feature).

5. Remove code tag

Press the × icon on a code tag to remove it from individual responses.

6. Add code tag

Press the + icon on responses to assign codes.

7. Save changes

Press Save to make all coding changes permanent and sync them to your Protobi project.

The advanced recode page offers significantly more editing capabilities. For a comprehensive guide on manually editing and applying codes, refer to the Recode (advanced) tutorial.

Protobi AI Autocode covers every stage of the open-end coding workflow — building a codeframe, tagging responses, and validating assignments. Each option is designed for a specific stage, so you can apply as much or as little automation as your project requires.

Related articles

• Coding Open-End Responses in Protobi →
• Simple Recode →
• Advanced Recode →
• Evaluate Coding Quality →
• Word Clouds →
• Verbatim Charts →

For questions or help, contact support@protobi.com.