A saved reply is written for one customer. Your support bot can send it to anyone.
A common way to give a support bot knowledge is to let your agents save their good answers. A ticket gets solved, the agent clicks save, and the answer joins a set the bot retrieves from when a similar question comes in. Many support tools offer some version of it. We built ours.
The article discusses the challenges of using saved customer support replies as knowledge for AI bots, highlighting issues that arise when these replies are reused without modification.
What happened
The article describes how a company discovered that their AI support bot was using saved customer support replies that contained sensitive information and specific context. They found that these replies needed to be rewritten to be generic and suitable for all customers. The company then implemented rules to ensure future replies are more appropriate for AI use.
Why it matters
For AI developers and customer support professionals, ensuring that AI systems use appropriate and generic information is crucial to avoid providing incorrect or harmful responses. This article highlights the importance of careful curation of knowledge sources for AI systems.
- Ensure that saved support replies are generic and free of sensitive information before using them in AI systems.
- Implement rules and checks to prevent the reuse of specific customer data in AI knowledge bases.
- Regularly review and update the knowledge base to maintain the quality and relevance of AI responses.
The bigger picture
The article emphasizes the importance of cleaning and curating data used by AI systems to prevent the reuse of sensitive or inappropriate information. This can lead to more reliable and safe AI support systems.
The article highlights the challenges and solutions for using customer support replies in AI systems, emphasizing the need for careful curation and rewriting to ensure safe and effective AI interactions.
Technical details
- Saved answers are retrieved based on cosine distance in embedding space.
- The AI bot uses embeddings to find similar questions and retrieve relevant answers.
- Rewriting answers involves removing specific customer data and making the content generic.
- The system checks for sensitive information before adding answers to the knowledge base.
FAQ
What is the main issue with using saved support replies in AI bots?
The main issue is that saved replies often contain sensitive customer data and specific context that is not suitable for reuse by AI bots.
How did the company address the problem of uncleaned replies?
The company rewrote the saved answers to be more generic, removed sensitive information, and implemented rules to ensure future replies are appropriate for AI use.
What are the key rules for creating reusable support answers for AI bots?
Key rules include avoiding customer data, not including account links, not describing actions taken, and ensuring answers are clear and cover all aspects of the question.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by Dev.to AI on Oct 14, 2026.
Comments
Sign in to join the discussion.