These are the posts about what went wrong, because that is where most of what I know comes from. The first WhatsApp automation I built for a hospital looked solid until real patients replied in ways the flow never expected. The fix was a fallback to a person at every point the bot could not be sure, and after that manual follow-up dropped by 80 percent with not a single enquiry missed.
Other lessons were smaller and more embarrassing. A scheduled job once ran twice and sent the same messages twice, so now every job I build can only run once at a time. A template was rejected for sounding like an offer. A project slipped because the scope kept growing mid-build, so I now agree one small first version before building anything. Each post here is one of those moments, and what I changed afterwards.
