We've been running AI voice agents in production for small businesses in Ireland — plumbers, electricians, dental practices, restaurants. Here's what we've learned about the trades vertical specifically.
The Problem Is Simpler Than You Think
Irish tradespeople miss 40-60% of incoming calls during working hours. Not because they're bad at business — because they're physically doing the work. You can't answer your phone when you're soldering a pipe.
The existing "solution" is voicemail, which has a ~5% callback rate in 2026. People just call the next business on Google.
What the AI Actually Does
For trades, the call flow is different from, say, a dental practice:
Caller: "Hi, my kitchen tap is leaking badly"
AI: "I'm sorry to hear that. Is the water actively flowing,
or is it a slow drip?"
Caller: "It's spraying everywhere"
AI: "That sounds urgent. I'm going to text [plumber name]
right now with your details. Can I get your address
and phone number?"
The key features:
- Emergency triage — distinguishes "burst pipe flooding my kitchen" from "I'd like a quote for a bathroom renovation next month"
- Detail capture — location, problem description, access info, preferred callback time
- Instant notification — urgent calls trigger immediate SMS to the tradesperson
- After-hours handling — takes calls at 11pm when a tenant's heating dies
Technical Decisions That Mattered
Latency over everything. Sub-500ms response time is non-negotiable for voice. We process in the media pipeline, not via external API round-trips.
Accent handling. Irish accents vary wildly between Dublin, Cork, and Galway. We needed robust ASR that doesn't choke on "How's she cuttin'" as a greeting.
Domain vocabulary. The AI needs to know that a "combi boiler" and a "combination boiler" are the same thing, and that "the immersion" is an Irish-specific term for a water heater.
Results
One plumber in Galway: missed call rate dropped from 55% to <5%. Estimated €2,400/month in recovered jobs. AI cost: ~€50/month.
The ROI is absurd because the bar is so low. You're not replacing a human receptionist — you're replacing an unanswered phone.
What's Next
We're expanding to more trades verticals (HVAC, landscaping, cleaning services) and building out the Spanish-language version for the Argentine market. The core insight applies universally: service businesses where the owner IS the workforce have the biggest missed-call problem and the highest ROI from AI answering.
Would love to hear from other devs building voice AI for SMBs. What verticals are you seeing the most traction in?
Top comments (1)
great writeup and the insight about the owner being the workforce is the key one. the missed call problem actually gets worse as the business grows because more jobs means less availability to answer, which is a brutal feedback loop.
one thing we found building in the US market is that the urgency triage piece you described is probably the most critical feature to get right early. callers with a burst pipe or no heat in winter dont want to walk through a structured intake, they want to know someone is being notified right now and will call back in minutes.
the after-hours use case is where the ROI becomes impossible to argue with. people calling at 11pm with a cold house dont Google alternatives in the morning. they call back whoever they spoke to first. whoever owns that conversation wins the job.
curious what youre seeing for false-urgent rate in production, where the emergency triage fires on calls that turn out to be routine when someone actually calls back.