Confidence and caveats

We remind early piloters of our AI bot that your confidence in the AI-generated responses should be provisionally downgraded based on the following factors:

• we have not yet added the full set of resources to our ‘walled garden’
• we continue to adjust the rules and prompts based on ongoing testing
• we have not yet fully explored key ‘human-in-the-loop’ (HITL) processes or tested a range of third-party large-language models (LLMs).

We also remind early piloters of our AI bot about the following caveats to the responses you receive:

• they may contain ‘hallucinations’ (e.g., fabricated resources or statements)
• they may make statements that cannot be aligned to content in our ‘walled garden’ of resources from sources.