.png){kind=logo}
what's the best way to do QA on AI generated emails?
what's the best way to do QA on AI generated emails?
We use AI to automate QA for customer support emails. Our system runs numerous checks, focusing on: 1. How well the question was solved 2. The tone of the response 3. Understanding of the customer's issue 4. Setting timeline expectations This approach allows us to provide quality assurance at scale, generating an average score across tickets and flagging specific ones for review in Slack. It's not used for performance management, but as a first line of defense to ensure a minimum quality standard. For a detailed walkthrough of this process, you can check out our full Claybook tutorial on automating QA for customer support tickets.
Andi D.··i want Claygent to do QA on finished email copies
While we don't have a specific feature for QA on finished email copies using Claygent, you can adapt our AI-powered QA process for this purpose. Here's how you could approach it: 1. Create a new table in Clay and import your finished email copies. 2. Use Claygent to set up custom outputs for QA criteria, such as: • Tone consistency • Grammar and spelling • Message clarity • Call-to-action effectiveness 3. Configure Claygent to analyze each email copy based on these criteria. 4. Run the process on your dataset to get structured QA output for each email. This approach allows you to automate QA at scale for your email copies. For a detailed walkthrough, you might want to check our Claybook tutorial on automating QA for customer support tickets, which you can adapt for email copy QA.
While we don't have a specific pre-made prompt for catching weird AI-generated emails, we can help you create one using our AI prompt writing features in Clay. Here's a general approach you can take: 1. Create a new column in your Clay table for QA checks. 2. Use the AI message drafting feature to create a custom prompt. 3. Your prompt could include criteria like: • Check for appropriate tone and language • Ensure content relevance to the recipient • Verify correct personalization • Flag any potentially offensive or inappropriate content • Assess overall coherence and readability You can refine this prompt based on your specific needs and the types of issues you're trying to catch. Would you like some tips on how to write effective AI prompts in Clay for this purpose?
