AI coding

AI coding is currently in beta and is not yet available to DIY users. This feature may incur additional cost - please talk to your Glass representative for more details. 

This feature will continue to evolve, with enhancements added over time. If you are interested in enabling AI coding or learning more about access and pricing, please reach out to your Glass representative.

AI coding helps you quickly organize and interpret open-ended responses by automatically generating categories and assigning responses to those categories. This feature is designed to accelerate early-stage qualitative analysis and make open-end data easier to filter, summarize, and explore. AI coding is available directly within Glass and does not require exporting data to an external tool.

Navigate to Analyze Results, then locate the Build Export section across the top of the page.

Select the AI Coding tab. You will see a list of eligible open-ended questions, each with a gear icon next to it.

Click the gear icon labeled Configure next to the question you want to code.

When you click Configure, a pop up opens with three tabs:

1. Categories

2. Classification

3. Summary
   

You will move through these tabs sequentially.

The Categories tab is where you define how responses should be coded.

You can select one of two coding approaches:

• Theme coding Designed for longer, open-ended responses. This option generates higher-level themes or ideas that may span multiple concepts within a single response.

• Coded list Designed for shorter responses or lists of words or phrases. This is especially useful for questions like unaided brand awareness, where respondents list items rather than explain them.
  

After selecting a coding type:

• Choose the number of categories you want to generate.

• Click Generate.

Glass will use its AI companion to review all responses to the question and generate the requested number of categories.

Each category includes:

• A short code

• A longer, descriptive category name

• A description explaining what types of responses fall into that category

Once generated, you can download a CSV of the categories and lock them if needed. At this stage, the file includes only the category definitions, not respondent-level coding.

You can also edit category names and descriptions or add your own custom categories to better reflect your research needs.

The Classification tab applies the generated categories to individual responses. Click Classify to begin processing. This step may take longer to complete, as a single response can be assigned to multiple categories.

Once classification is complete:

• Each respondent’s response appears alongside category codes showing how it was categorized.

• Responses that were not categorized are clearly indicated.

You can:

• Download a CSV containing respondent IDs, responses, and assigned categories

Important: This classification output does not save automatically. Be sure to download the CSV, as it contains the respondent-level data needed for further analysis.

The Summary tab provides a high-level view of your coded results. You will see:

• Each category

• The number of responses assigned to that category

• The percentage of total responses represented

The most actionable output from AI coding is the classification CSV.

This file includes respondent IDs associated with each category, allowing you to:

• Build filters based on category membership

• Create respondent ID lists

• Use those filters in Crosstabs, Charts, or other analyses

For more details on working with respondent IDs, see the Building Respondent ID Filters article.