An AWR report is a sampling of a moving system, not a verdict. The first thing we do is confirm the snapshot window actually covers the complaint. A one-hour window that straddles a quiet period will flatten the spike you were asked to investigate.
Next we anchor on database time per second rather than raw wait counts. If DB time per second is close to the number of active sessions, the instance is saturated and no amount of index tuning will rescue it until concurrency is addressed.
Only then do we look at the top timed events, and we read them alongside the SQL ordered by elapsed time section. A single statement responsible for 60 percent of elapsed time turns an architecture debate into a one-afternoon fix.
Finally, we compare against a known-good baseline. Without a baseline you are guessing at what normal looks like, and every report will appear alarming.