Smart Narratives are very useful for systematically creating dynamic summaries in Power BI, but they have a few limitations:
Limited Types of Visualization Support: Smart Narratives work brilliantly well with visuals like cards, bar/column charts, and line charts; there's little hope for them in more advanced visuals such as scatter plots, maps, matrices, or custom visuals to produce good insights.
Static Snapshot: Although Smart Narratives are dynamic because they update with filters or slicers, they do not support drill-throughs, bookmarks, or advanced interactivity. They provide high-level summaries but cannot narrate more granular or user-driven exploration.
Customization Constraints: You can edit the text and add your measures to the narrative, but there is limited flexibility in customizing the logic behind auto-generated insights. In fact, you really do not have control over how Smart Narrative interprets data patterns.
Performance Notes: Smart Narratives will be able to lower performance on reports that feature larger datasets or many complex measures, particularly while filtering or recalculating.
It's quite similar in short form. Although Smart Narratives lend themselves most to storytelling, they are best enhanced by the presence of classical visuals and do not replace that need with more profound analytical exploration.