Hot take: Drata's chatbot support is worse than just having a knowledge base

Just spent the last 45 minutes trying to get a clear answer from Drata's in-platform chatbot about a specific evidence mapping rule for SOC 2. I came out of it feeling like I'd have been better off just CTRL+F'ing their docs, which is... not the point of an AI assistant.

Here's the thing: the chatbot seems trained on generic Drata marketing and high-level FAQ content. When you ask it something technical—like, "What's the exact path for the `auth.log` file evidence collector on an Ubuntu 22.04 system for the 'Failed Login Attempts' control?"—it either gives a vague paraphrase of a help article or confidently states something that's subtly wrong. It lacks the deep, contextual understanding of the platform's own logic and configurations.

For example, I tried:

```
My question: Can I customize the frequency of automated evidence collection for a specific policy?
Chatbot response: Yes, Drata allows for flexible evidence collection schedules to meet your compliance needs. You can adjust settings in the Policy section.
```

That's not an answer. That's a deflection. The *actual* steps involve navigating to a specific control, editing the evidence collector, and toggling a schedule dropdown—details the bot either doesn't know or won't share.

It feels like a step backwards. A well-organized, searchable knowledge base at least gives you definitive (if static) information. This chatbot gives you the illusion of a conversation while often sending you on a wild goose chase. I'm all for AI integration, but this implementation feels more like a checkbox feature than a real tool.

Has anyone else run into this? Found a prompt hack that makes it more useful, or are we all just quietly going back to the docs and support tickets?