Applies to Product- Power Automate
What’s happening?
The AI Builder prompt is not consistently returning accurate results when querying a Dataverse table of acronyms, leading to only a few of the 800+ acronyms being recognized.
What’s happening?
The AI Builder prompt is not consistently returning accurate results when querying a Dataverse table of acronyms, leading to only a few of the 800+ acronyms being recognized.
Reason:
The underlying cause appears to be the inherent variance associated with large language models (LLMs), which may affect their ability to reliably consume the acronym table and return the expected definitions and examples.
Resolution:
To address this, it is recommended to implement a Power Automate flow that performs the acronym lookup in the Dataverse table. This flow should provide the meaning of the acronym as its output. The output can then be integrated into an adjusted version of the existing prompt to enhance the overall functionality. Additionally, it may be beneficial to discuss handling responses that do not match any entries in the table and suggesting alternative acronyms that the user may have intended.