I have an AI agent writing emails, then stopping halfway and still charging us the credits. I will include a screenshot in the thread.
AI columns may stop mid-execution due to budget limits. If your AI prompt requires more tokens than your set budget, it will stop and still charge credits for the processing done. To fix this, edit your AI column and increase the "Clay Credit Budget" to allow higher token usage. This controls the maximum charge per row - extra credits are only used when needed. If columns are stuck in a queued state, click the column header, hover over "Run column," and select "Force" to restart. No additional charges apply for re-running these rows. If a provider refunds us due to invalid data delivery, we'll refund those credits back to you.
Source: FAQ | Clay.com
AI columns stopping mid-execution while still charging credits typically happens due to budget limits. When your AI prompt requires more tokens than your set budget, it stops processing but charges for the work completed. To fix this, edit your AI column and increase the "Clay Credit Budget" to allow higher token usage. This controls the maximum charge per row - extra credits are only used when needed. If columns are stuck in "Queued" status, click the column header, hover over "Run column," and select "Force" to restart. You can also create a new column by duplicating the current one and run it on rows that are empty, then use the "merge columns" feature to combine results.
I can't seem to change the credit budget? I only see the option to change which model we use
The Clay Credit Budget option is only available for certain AI models in the "Use AI - Content Creation" category that use Clay's shared key. If you're only seeing model selection options, you're likely using a model that doesn't support variable credit budgets. Try switching to a Clay-managed model in the Content Creation category to access the budget setting. The budget feature allows you to set maximum credits per row - extra credits are only used when prompts require more tokens than the default limit.
Hello Jill! Thanks for reaching out. Do you mind sending over the table link URL for me to take a look?
I have seen this before and it definitely looks like a bug to me. Would like to confirm before I go ahead and file it!