couple months ago a "client's" whatsapp agent stopped reaching anyone. my n8n workflow showed every message as sent, because the cloud api still answers with a message id the status webhook told the real story: failed, error 131031, "business account has been locked". meta had locked the client's business portfolio. the whatsapp account still showed active in the dashboard i lost an evening blaming my access token first. if the status webhook had been wired from day one, i would have known in a minute no code change fixed it. we moved the number to twilio as the provider and it was back up the same day. and now every workflow has a status callback that alerts me on telegram when a message fails for those running whatsapp for clients: are you on meta directly or through a provider like twilio, and why?

someone im connected with on linkedin booked 30 min through my calendar link. he helps run a medical clinic so honestly a real fit for a voice agent. i spent the morning preparing my questions 9:30 im on the meet. nobody at 9:31 i emailed him that i was on the call. at 9:50 i sent a second one with a new time later this week, same link. still no reply maybe he just forgot, maybe something came up and he really couldnt. either way i wasnt in a great mood the rest of the day next time im sending a reminder the evening before with one line on what we'll cover, so he has a reason to show up or a chance to move it how do you guys cut no-shows on discovery calls?

Last month I found a hole in the intake prompt of the phone agent I demo to plumbing and HVAC shops. It collected name, address, callback number and the problem, then sent the call details to an n8n workflow. Fixed keywords like "gas smell" or "sewage backing up" flagged the call as urgent so the owner got a text first. Then I went through an open-source intake template for contractors and saw they handle danger before anything else. If the caller mentions a gas smell or a CO alarm, or anything electrical that's sparking or wet, the agent stops and tells them to get out and call 911. Before it asks for a name. Mine would have asked someone smelling gas to spell their street name. I added it as step 0 and checked the live prompt through the Retell API afterwards. If you build intake flows, voice or chat, what do you handle before collecting any details?

for plumbing prospects i stopped showing a generic demo. each shop gets a page at /demo/their-name and the agent answers with their company name setup is simple. its one Retell agent with business_name and trade as dynamic variables. the demo page passes them when it creates the web call the part i almost got wrong: if the page just reads the name from the url, anyone can type whatever and the agent will say it out loud. so the server has a small allowlist that maps the slug to the real name, anything not on it gets rejected also keep the demo agent separate from your production one. one bad edit and a real clients line picks up with the wrong business name has anyone done this at bigger scale, like generating the demo page automatically from a lead list?

Before pitching a phone agent to a plumber, I wanted to see how plumbers themselves talk about missed calls. A Plumbing Zone thread changed how I'd pitch it. The guy who started it runs a 3-person shop: 47 missed calls in a month, 5 voicemails. He'd tried his wife answering and a $1,500 a month receptionist. The replies from other plumbers were blunt. One misses 5 to 10 calls a day and says most are price shoppers. Another tagged every missed call for a few weeks and found most of them were junk. So the first thing worth automating might just be the counting. Log each missed call and tag whether it was a real job, before anyone sells the owner an agent. If you've built for trades or home services: did you measure the missed calls before building, or after the client asked why the numbers didn't move?