@Deshmukh, Ashish For this scenario I would consider using the custom NER project to train a custom model to extract entities that are specific to your organization.
For ex: In this case the number "23456" is an employee number and you would like the endpoint to extract and label this data in the sentence as employee number.
To achieve this, you will have to train a model with some sample sentences and label the entities. This can be done using the language studio as a visual interface or you can create a formatted labels JSON file and import it in your project. A sample project that you can use is available to test on github. You can use this quickstart along with this sample and test the deployed endpoint to extract the entities.
You can also use the readily available NER API to extract some supported entities, but it would not identify an employee number as employee number as this is not yet supported.
If an answer is helpful, please click on or upvote which might help other community members reading this thread.