Won a hackathon where we wrote test cases into a spreadsheet in plain human language. The tool would make a ticket in Jira, write the test, run the test to make sure it fails (because the feature hasn't been implemented), link it with Buddy, makes a pull request in GitHub, and notifies project owners on Slack.

The backbone of this was Pipedream, Jira, and Buddy, and GPT-4 handled the test writing. This was a year ago when GPT-4 was the top of the line AI model. Yes, we needed all of these, it was a fragile thing built on external tools.