Prerequisites
Before you run a New Content Generation job:
Make sure you have created an Aisera application or bot and added a Data Source to it.
Review the Knowledge Base Article Generation Policy for your bot, and choose an AI model for your generation job (you might need your Aisera team to do this for you - depending on permissions).
Click View Policy under the Knowledge Generation Policy section. A configuration window opens, displaying the prompt settings used for KB Article generation.
From the dropdown, select the supported AI model based on your agreement during onboarding or sales discussions.
Supported Models Include:
GPT-4o
LLaMA 3.3
LLaMA 3.1
When you choose a Data Source for your application from the Aisera Admin UI, the fields to transfer data from the Aisera Gen AI platform to your data source have already been mapped. Review and customize the field mapping for your application, as needed.
Make sure you have a minimum of 40K tickets to form a homogeneous cluster.
It is essential that you review the tickets before passing them through the Knowledge Generation extraction model. The quality of the resulting Knowledge Document is determined by the quality of the Tickets you provide.
Use the Content Generation > Knowledge Generation > Actions > Condition window to choose the correct data source for ticket consideration when you're generating a KB Article.
You must select a Data Source under Conditions before creating the KB Generation job.
For more information, see Pre-Generation Parameters.
Make sure to pass the tickets that have good-quality comments (which actually talks about resolution, we saw many tickets have comments like ‘Issue Resolved”, “Completed”, ‘Internally we fixed the issue” etc, which will not be useful to generate the Knowledge) Also, the comments should be in sync with the description and title. During our investigation, we found that some tickets had titles, descriptions, and comments that were not related. Tickets like these will generate a low-quality knowledge base (KB) that is of no use.
Select the tickets that are marked as Closed or Resolved. This selection in the Conditions (configuration) is based on the assumption that closed and resolved tickets tend to have valuable comments that provide meaningful resolutions. However, it's important to note that our system has the capability to scan and extract meaningful resolution notes from all tickets, not just tickets with closed or resolved status.
Ensure that all relevant comment fields are ingested from the source system to Aisera. KB Generation considers comments to be of utmost importance, so it's crucial to include all types of comments present in the source system. Keep in mind that each customer may maintain different fields to store their comments, so it's essential to understand these fields and ingest them accordingly
Once the comments are ingested, it is crucial to map various comments fields to the CaseComment text field in the Data Source from where you ingested the tickets
See Field Mapping or see your Aisera contact for more details.
Select an Existing Job
If you are running an existing job, specify the Job Name and the kb-gen-job-id, data-source-id, or bot-id.
Last updated