Release notes - 3.5.3

Release date: 27/11/2025 (after workinghours)

Improved unanswered questions check for generative scenarios

We’ve improved how Oswald identifies truly unanswered questions in generative scenarios. This update helps better flag cases where a user received a response from the LLM, but their underlying question or problem was not actually resolved.

• How it works: Oswald now uses a stricter prompt in its second LLM call to assess whether the user’s question was genuinely answered. You can review flagged items via Train > Unanswered Questions.

What this means for you

If you’re using generative scenarios, this update improves how Oswald flags cases where your chatbot didn’t really help the user despite answering. This does not affect non-generative chatbots or clients not using this feature. No action is needed on your part. Simply continue reviewing flagged questions via Train > Unanswered Questions.

Need help or want this feature enabled? Feel free to reach out.