Wat kan er misgaan met agent-gegenereerde dashboards?

Samenvatting

what could go wrong with agent-generated dashboards? we’ve been playing with generating dashboards from natural language instead of building them manually. you describe what you want, it asks a couple of follow-ups, then creates something. on paper it sounds nice. less time on UI, more focus on questions. but i keep thinking about where this breaks. data is messy, definitions are not always clear, and small mistakes in logic can go unnoticed if everything looks clean in a chart. also not sure...