Any Famulor Pros here?

Help & Questions
1

Currently building an AI voice receptionist for a veterinary clinic in Famulor.

The booking flow works well overall, but after reviewing real clinic test calls I found a major issue:

  • emergency/urgent calls are being handled too dynamically by the LLM
  • sometimes calls end too abruptly after routing attempts
  • trust drops immediately if the flow feels uncertain

After reviewing the calls, I think the correct solution is moving more logic into an IVR menu before the AI conversation starts.

Example idea:

  • Press 1 for bookings
  • Press 2 for general questions
  • Press 3 for urgent/emergency cases
  • Press 4 to talk to staff

Then each path would have a more controlled flow instead of relying on the LLM to classify intent mid-call.

My questions:

  • Has anyone built a similar IVR + AI flow inside Famulor?
  • What’s the best way to structure this?
  • Any best practices for handling urgent/emergency calls without abrupt endings or confusing routing?

Would appreciate any guidance from people who’ve built production voice systems in Famulor.

From Skool · Original date label: Mai 19 • :light_bulb: Feature Requests

2

you’re spotting exactly what breaks in production voice systems: high-stakes intent (emergency) + open LLM routing + unclear handoff = trust loss.

Your IVR-first instinct is right. In Famulor, the most reliable pattern is usually:

  1. Deterministic menu first, AI second: Use Flow Builder and put a keypad collection step at the very start (Collect keypad input):
  • Press 1 → Bookings (tight, scripted flow)
  • Press 2 → General questions (lighter AI, strict KB)
  • Press 3 → Urgent/emergency (almost no improvisation)
  • Press 4 → Staff (transfer path)

Then branch into separate subflows per option instead of one big prompt where the model classifies intent mid-call.

  1. Keep emergency logic out of “free conversation”For option 3, avoid dynamic LLM decisions like “is this urgent?”Use a fixed script:
  • acknowledge urgency in one sentence
  • collect only minimum data (pet name, symptom severity, callback number)
  • always explain next step (“I’m connecting you to our on-call team now, please stay on the line.”)
  • then warm transfer (preferred) or cold transfer

Also enable “Send message before transfer” so callers never experience a sudden hangup.

  1. Why calls feel abrupt today: Common causes we see in clinic setups:
  • transfer triggered without a closing phrase
  • LLM ends call after failed routing attempt
  • no fallback when transfer fails (should route to voicemail + ticket/SMS, not end)

Add explicit fallback nodes:transfer failed → apology + alternate number + voicemail/task creation.

  1. Structure that works well in Famulor
  • Inbound assistant #1 (main line): menu + routing only
  • Branch flows: booking / FAQ / emergency / staff
  • Emergency branch: deterministic + warm transfer to on-call
  • Booking branch: can stay more AI-driven (calendar tool, confirmations)
  • Staff branch: immediate warm transfer with short briefing to staff

Note: one phone number supports one inbound agent, so use Flow Builder branches (or external PBX IVR routing to different numbers/assistants), not four separate inbound agents on the same number.

  1. Best practices for urgent/emergency
  • Never rely on LLM intent detection alone for triage
  • Always use a “next step” sentence before transfer/end
  • Prefer warm transfer for trust (“stay on the line”)
  • Add retry for keypad (invalid input → repeat menu twice → default to staff)
  • Add clinic disclaimer: “For life-threatening emergencies, come in immediately / call local emergency vet”
  • Test with recorded calls per menu option (especially 3 and 4)
  1. Optional hybrid (very robust)Some clinics run a classic PBX IVR before Famulor (Press 1/2/3/4), then forward to different Famulor assistants/numbers. That gives maximum control if you already have telephony infrastructure.

If helpful, share your current flow (booking vs emergency node) and I can suggest a concrete Flow Builder structure for the vet clinic case.

From Skool · Original date label: • Mai 22