Applying Distinct AI Models to Subsets of Dataset Rows

Yes, it's possible to apply different AI models to specific subsets of rows in your dataset using Clay's conditional logic and AI features. Here's how: 1. Use AI formulas to set conditions for running specific models on certain rows. 2. Implement conditional runs to ensure AI models only process rows meeting your criteria, saving credits and improving efficiency. 3. Choose from various AI models like Claude Sonnet 3.7 or Google's Gemini 2.0 when setting up your "Use AI" enrichment. 4. Configure the "Reasoning Budget" for Claude models to control capabilities for specific subsets. By combining these features, you can tailor AI processing to different parts of your dataset based on your specific needs.

Sources: Flexibly run any workflows based on conditional logic | Clay AI Formulas + Conditional Runs [Clay 101] - Clay University Credit Reporting, X.com source, new Claude and Gemini models, Instagram updates, and more!

To change the AI model for rows that failed after using ClayAgent and prompt engineering, you can follow these steps: 1. Go to the column where the AI enrichment was applied. 2. Right-click on the column and select "Run Column" > "Run All Rows that haven't run or have errors". 3. In the AI enrichment settings, click on "Use AI" and choose a different model (e.g., "Claygent Neon" or another available option). 4. Adjust your prompt if needed. You can use the "Help me" button for assistance in refining your prompt. 5. If you encounter budget limit messages, consider increasing the "Clay Credit Budget" in the AI column settings to allow for more complex prompts. If you need further assistance, please request to talk to a person.

Sources: Metaprompter, shared workbooks, new hire signals, Outreach integration, and more! Writing AI prompts in Clay | Documentation | Clay University Run Options

Yes, you can run a different AI model for selected rows using conditional runs. Here's how: 1. When setting up your AI integration, go to "Run Settings" at the bottom of the configuration panel. 2. In the "Only run if" box, create a conditional formula to specify when the AI should run. This allows you to target specific rows based on certain criteria. 3. You can use the "Use AI" button to help generate the formula if you're not familiar with the syntax. Alternatively, you can use filtering: 1. Filter your table to show only the rows you want to process. 2. Right-click the AI column and select "Run Column" > "Run All Rows that haven't run or have errors". This approach helps save credits by running the AI only on relevant rows.

Sources: Run Options Guide: Ways to save Clay credits | Documentation | Clay University Use AI integration overview | Documentation | Clay University Conditional Formulas

To change the prompt for a single row in Clay: 1. Open the "Use AI" enrichment for the column you want to modify. 2. Click on the "Prompt Editor" to access the configuration settings. 3. In the editor, you can customize the prompt for individual rows by using dynamic inputs. 4. Link the appropriate columns to the dynamic prompt inputs to ensure row-specific data is used. 5. Use the "Required to Run" toggle for fields that must contain data before the enrichment runs on a specific row. 6. Preview your results by clicking on any enriched cell to view the generated details. Remember, you can always modify existing prompts to suit your specific needs for individual rows.

Sources: We Made 24 Prompts So You Don’t Have To [Automated Outbound] - Clay University Writing AI prompts in Clay | Documentation | Clay University Referencing dynamic data from other columns | Documentation | Clay University

Hey there, thanks for reaching out! 😊 That should be possible, but be careful not to run it for all rows. If you change the model and then click "Run all rows that haven't run or have errors," it will rerun everything since the column was modified. You have two options: 1. Change the model, filter for failed rows, and only run those. 2. Create a second column and run it conditionally on the failed ones with a different model. Let me know if you have any other questions!