AI Agent 5 second delay on responding

[steve.moran]

I am testing an AI Agent using OpenAI to act as a receiptionist. I have uploaded a very basic json file of branch numbers to the Knowledge base, but when I call in as a caller and interact with it, there is a very noticeable 5 second delay each time I ask it a question before it responds. It works fine but the delay is uncomfortably long and a barrier to actually using it.

I have tried all the options available gpt5.4 and gpt5.4-mini but the delay is still there.



Is this just normal behaviour for how 3CX has implemented this, or do I have some configuration wrong?

Not sure these are relevant since they say they are post-call actions but here are my transcription settings.



The delay is a real barrier to adoption of this service as customers are just not going to put up with it.

Thanks

Steve

 

[TheodorosG_3CX]

Is there any knowledge base used for the AI Agent?
It is also really important that the knowledge structure is configured properly. Check this guideline also to ensure that it will always perform at it's peak.

 

[Kevin Attard Compagno]

In addition, I would convert the JSON knowledge to an MD file, AI is typically happier with that.