Hi all, software engineer here with no ML background.

I’m thinking to build a support, to help engineering teams mitigate time spent in issues.

I was thinking to apply some ML and have the model learn over time, and default to “create a support ticket” If it doesn’t have enough accuracy for an answer (or if the user rated it as not helpful).

At first, the bot will be just creating tickets, but over time, I was thinking to have it learn by the different resolutions the engineers give (ie. Link to documentation, grant access to a system, etc).

Is this even a possibility? Having a machine learning model untrained and have it learn over time? Any other suggestions?

I could have some initial data, but I’m not interested in súper old data, as the systems and documentation may have changed