hitting error on like 1/4th of results with a message: An unknown error message - while running ai columns - claygent and local ones
That's frustrating—let's get this sorted. Try these quick fixes:
First, clear your browser cache:
Windows: Ctrl + Shift + R
Mac: Cmd + Shift + R
If using OpenAI API columns, check your API key:
Verify it's correct at OpenAI API Keys
Update in Clay if needed
For rate limit issues:
Check if you're hitting API tier limits (need 30k TPM for content generation, 450k TPM for Claygent)
Monitor which specific columns are failing most
Click into the specific cells showing errors to see more detailed error messages—this will help pinpoint the exact issue.
Which of these steps helps, or what specific error details do you see when clicking into the failed cells?
If you still need help resolving the errors with your AI columns, I’m here to assist. Would you like to provide more details about the specific error messages or the steps you’ve tried so far?
same 👋
Got it—still seeing those "unknown error" messages on 1/4 of your AI columns.
Try Clay's new "Troubleshoot with AI" feature:
Click any failed cell
Select "Troubleshoot with AI"
Get instant debugging instructions tailored to your specific error [1]
Also check if you're hitting rate limits—recent Claude Sonnet updates doubled token usage, which can cause more failures on lower API tiers.
If you're using OpenAI columns, verify you have 30k TPM for content generation or 450k TPM for Claygent.
What does the troubleshoot feature show you?
If you still need additional help with the errors in your AI columns, I’m here to assist. Would you like to share more details about what you’re trying to resolve or any other issues you’re encountering?
Correct—this isn't specific to your account. The "unknown error" messages affecting 1/4 of AI columns are related to recent Claude Sonnet updates that doubled token usage, causing more rate limit issues across the platform.
We're actively deploying fixes like retries and smarter pacing to handle these rate limits better and reduce errors system-wide.
Is that what you were looking for?
no, i use open ai right now with the issue
Thanks for clarifying. I’ll connect you to a human agent who can assist further with your OpenAI issue.
Our support team has got your message and we'll get back to you soon!
If you’re dealing with a specific table, drop the URL below so we can help you quicker. Otherwise, someone from our team will be in touch soon!

Hey there! Do you mind sharing a table URL link so we could check it out further!
It looks like many of the AI columns to the left of the table successfully completed (though you may experience slower return times due to having a large amount of rows) but the ones to the right such as 'Normalise First Name' have a combination of returned values, run condition not being met, or errors. With the run condition, it looks to be that the columns you're dependent on are empty or edited. With the errors, I believe if you change the prompt it could work better.
I recommend throwing it into ChatGPT and asking it to re-prompt it (accordingly to your desires). You can then paste it, make minor edits to details and re-reference your dependent columns.
Try this out and let us know if it works!
