I would like to discuss a task similar to “A convincing answer still needs a proper check.” A possible distinction: Requirements are clear, but errors appear unpredictably. Possible first output: Evaluation findings with examples of errors, limits on use and questions requiring technical investigation. I would like to check whether this first step fits our task and discuss terms for further work. Our situation: [add details] What we want to change: [add details] What we have tried: [add details]