Refine with heuristic evaluation
Heuristic evaluation was covered in Chapter 9, Guidelines and Heuristics. The same approach, along with classifying and user scoring, from Chapter 4, Scoring Stories, can be used at this stage of the development process. All of that applies to monitoring and evaluation results. We can classify issues, score them to prioritize improvements, then apply refinements to the data, prompts, and model improvements (or even test against a new model to merge into the solution), or work on a new integration. You’ve got this! Apply what you know. This leads us to a discussion on handling the process of conversational AI within typical development organizations in Chapter 11, Process.
