Hey team what's the best way to conduct quality control for AI emails? tried to use a formula like this to catch occasional AI delusions where instead of double spacing it outputs "\n\n" as text in the email, but it's not catching the mistakes
To catch formatting issues like "\n\n" appearing as text in AI-generated emails, you can use Clay's "Remove Extra Whitespace from Text" formatter: 1. Create a new column using the Clay formatter integration 2. Select the "Remove Extra Whitespace from Text" action 3. Input the text column containing your AI-generated emails 4. Run the enrichment to clean up the text Alternatively, you could use an AI formula to specifically target and replace escape sequences: 1. Create a formula column 2. Write a prompt like: "Replace all instances of '\n\n' text with actual line breaks in this text" 3. Preview the results to ensure it's working correctly For more complex formatting needs, consider using HTML tags with an AI formula that adds proper paragraph formatting (like `
Hey there, thanks for reaching out! This depends on whether you're referring to quality control in terms of content (like does this sound like AI or not?) or formatting (are \n\n's included or excluded?) If you're referring to content, we could take samples of 5-10% of emails and review them manually to have confidence in the AI or provide specific style guides that give examples of how you write emails. If this is in reference to formatting, it would be best to understand which email sequencer you're using and what would be the most consistent way to get double spaces into that email sequencer. The character "\n" represents a line break that most sequencers respect, but we'd have to test and make sure this is the case. Let me know if this helps!
thanks Owen! totally understsand that \n is a line breaker but 99/100 emails generated inside of Clay came out with proper spacing but then there's that 1% of emails just randomly came out with "\n\n" in text
and im using Instantly, when those are pushed into Insstantly they remain text
if you check the screenshot I sent earlier, only "William"'s records contain \n\n, that's why it's so odd
lol
so Im using Formula to disqualify subject title/email body that contains \ or \n
Ah gotcha, totally hear you when it comes to consistency of output. Wondering if it would be worth adjusting filters to try and account for just the "\n\n" instead of anything with "" or "\n". Happy to help dive in and troubleshoot this formula. Two things: 1. Do you mind sending the link (url) to the table so we can take a look? 2. Could you pick a few row numbers where this error is occurring so we can dive in and get this solved as quickly as possible? Excited to help!
Hey Owen, here's the table link: https://app.clay.com/workspaces/449716/workbooks/wb_0swlka2cuvNSD7eDVhN/tables/t_0swmj0h5jKFttD4nrXo/views/gv_0swmj0i9hXPbwegmgDr See Column "Needs Troubleshoot" and row 28 as an example. I also tired "\n\n" in formula but it auto turns into actual white spaces lol won't maintain it's text format so i coudlnt do it
and row 119
it's just tough because it happens here and there randomly. If I flag them and rerun the row, it AI would spit out the correct version
Hey, having that "Needs Troubleshoot" flag is a great move - it definitely helps catch and rerun inconsistent AI results. What you're seeing is likely caused by the prompt generating multiple answers and it's leading to format inconsistency, which can lead to those random issues. That's why rerunning sometimes gives a clean output. Looks like a model hallucination due to larger prompt here.
.png){kind=logo}
