19.5.1. AI Prompt Reporting
- 19 Sep 2024
- 2 Minutes to read
- Print
- DarkLight
19.5.1. AI Prompt Reporting
- Updated on 19 Sep 2024
- 2 Minutes to read
- Print
- DarkLight
Article summary
Did you find this summary helpful?
Thank you for your feedback
All AI response data is stored within Salesforce and can be reported on.
For each successfully analyzed call a new object in Salesforce is created called ‘Insight AI Prompt’ which stores the results received back from the AI for every active prompt which was run against that call.
Then for each prompt that has been configured a ‘Insight AI prompt config’ record will be created that stores the name and details for that particular prompt.
Relationship table
.png?sv=2022-11-02&spr=https&st=2024-10-18T05%3A22%3A18Z&se=2024-10-18T05%3A32%3A18Z&sr=c&sp=r&sig=AuQQ9oXEK0dQE8pUTgBRSsuXJFqpAJoTXcwr8IasAao%3D)
Insight AI Prompt
AI Prompt Configuration | The ID of the prompt configuration object |
AI Prompts Name | The name of the prompt |
Created By | The API user |
Insight | The related master Insight record |
Last Modified By | The API user |
Natterbox User | The user associated with the call |
Queue Name | The associated queue the call routed via (if any) |
Rating | The numerical value out of 10 that the AI responded with, in the event of an error this field will be blank (recommended for reporting) |
Reasoning | The reasoning response that the AI gave for the rating |
Result | The result from the AI, typically this will be the numerical rating out of 10, but in the event of an error the full string will be populated here |
UUID | The UUID of the prompt |
Insight AI Prompt Configuration
Active | Whether or not the prompt is enabled for processing |
AI Prompt Configuration Name | The numerical configuration reference |
Created By | The user who created the prompt |
Description | The user defined description for the prompt |
Explanation | Whether or not an explanation should be provided (only relevant for the Rating prompts) |
Id | Salesforce record ID |
JSON | The configuration in the form of JSON |
Last Modified By | The user who last changed the prompt |
Name | The user defined name of the prompt |
Owner | The owner of the prompt |
Prompt Contents | The user defined prompt itself |
Type | The type of prompt configured (i.e. Rating) |
So in order to report on the results for a particular prompt you must :
Link the Insight AI prompt object withe Insight AI prompt config object
Filter on the right scope of Insight records
Filter on the name of the prompt you wish to report on using the Name field from the ‘Insight AI prompt config’ record
Note: If you later change the name of your prompt you will need to ensure you amend your reports and dashboards to match it, otherwise they will fail to return any results.
Additional filtering
Additionally you can then filter your AI response data further by using the fields from the Insight record itself, these include :
Time of the call
The associated Account
The associated Contact
The associated Lead
The User/Agent which serviced the call
The Call Queue which serviced the call
If you need to filter on more technical data points on your calls (such as the PBX group that serviced the call) then you’ll need to form a relationship in your reporting to the Call Reporting Object and source those fields from there.
Was this article helpful?
