Building Voice AI for Vertical SaaS: Lessons from Vet Clinics and Auto Shops

Two industries that seem completely different — veterinary clinics and auto repair shops — have an almost identical phone problem:

• Staff are physically unable to answer (hands in surgery / hands in engine)
• High call volume (80-150/day vets, 40-70/day garages)
• Callers need triage, not just message-taking
• Missed calls = direct revenue loss (€6K-15K/month)

Building AI receptionists for both taught us some interesting things about vertical voice AI.

Domain-Specific Intent Classification

Generic NLU doesn't cut it. "My dog ate chocolate" needs emergency routing. "There's a grinding noise when I brake" needs service booking + urgency assessment.

We built vertical intent classifiers that understand:

• Vet domain: Emergency triage levels, species-specific knowledge, medication refill patterns
• Auto domain: Service types, parts inquiries, NCT/MOT prep, breakdown urgency

The Calendar Integration Problem

Both verticals use wildly different practice management systems. Vets use systems like RxWorks, Provet, or paper diaries. Garages use Autowork Online, TechMan, or whiteboards.

The abstraction layer matters more than the AI:

CalendarAdapter
  ├── RxWorksAdapter
  ├── ProvetAdapter  
  ├── AutoworkAdapter
  ├── GoogleCalendarAdapter (fallback)
  └── SimpleSlotManager (for paper-diary shops)

What Surprised Us

1. After-hours calls are 30-40% of total volume in both verticals. Completely unserved market.
2. Callers prefer AI to voicemail — 85%+ completion rate vs. 38% voicemail message rate.
3. The "describe your problem" call is where AI shines — structured data extraction from unstructured speech.

---

Working on voice AI for vertical markets? Happy to compare notes. We're building VoiceFleet.ai.