The Challenge: After-Hours Calls Going Unanswered
Prairie Comfort HVAC, a Winnipeg-based contractor with 8 technicians, faced a common problem: emergency calls arriving after 5 PM or on weekends went straight to voicemail. During harsh Winnipeg winters, when residential heating failures spike, missed calls meant lost emergency jobs—and angry customers calling competitors instead.
The owner, Marcus, employed one full-time receptionist at $4,500/month CAD. Even with this investment, evening and weekend coverage remained a gap. Most calls came between 6 PM and 10 PM, when Marcus was managing job scheduling and customer follow-ups in the field office.
In winter 2025, Marcus estimated he was losing 8-10 emergency jobs per week to voicemail. At $1,700 average job value, that represented $13,600-$17,000 in lost monthly revenue.
Why Mihron AI's Maya
Marcus evaluated three solutions: hiring a second part-time receptionist (cost $2,500/month), a traditional IVR system (inflexible and frustrating for callers), and AI-powered voice agents. He chose Mihron AI's Maya because:
- Natural conversation—callers don't know they're talking to AI
- Emergency detection—automatically detects "emergency" and cold-weather keywords like "-30°C"
- Direct appointment booking—books available emergency slots in Cal.com in real-time
- Escalation to human—transfers to a live tech if customer prefers
- Canadian compliance—PIPEDA-compliant encryption, audit logs, 90-day PHI retention
- Cost—$399/month CAD (Growth plan), no setup fees
How Maya Works for HVAC Emergencies
When a call arrives after hours:
- Maya answers: "Hi, this is Maya from Prairie Comfort. What's happening with your heating today?"
- Caller describes issue: "My furnace stopped working and it's minus 28 outside."
- Emergency trigger: Maya detects the cold temperature and word "stopped." She offers immediate emergency booking.
- Booking: Maya checks the tech schedule in Cal.com and books the next available emergency slot. For Prairie Comfort, typical emergency response is 2-3 hours in winter.
- Confirmation SMS: Customer receives SMS confirmation with tech name, arrival window, and callback number.
- CRM sync: Call details, customer info, and appointment sync to the team's job management dashboard automatically. Marcus's team sees the new emergency job the moment the call ends.
If the customer prefers a live person, Maya transfers to Marcus's on-call tech or office staff. About 8% of calls request escalation; the rest complete booking with Maya.
Results: 3-Month Performance
| Metric | Before Maya | After Maya (3 mo) | Change |
|---|---|---|---|
| After-hours calls/week | 28 | 92 | +229% |
| Emergency jobs booked/week | 3 | 11 | +267% |
| Emergency jobs (3-month total) | 12 | 47 | +35 jobs |
| Emergency job revenue | ~$18K | ~$80K | +$62K |
| Dispatch callback time | 5-8 min | 30 sec | -95% |
| Staff phone time/week | 28 hrs | 8 hrs | -71% (20 hrs freed) |
| Monthly operating cost | $4,500 | $399 | -$4,101/month |
Key insight: The largest win wasn't the cost reduction (though $4,101/month savings is significant). It was the 35 additional emergency jobs. Marcus's tech team went from scrambling to fill dispatch slots to having a consistent stream of emergency calls—and happy customers who got immediate response instead of voicemail.
Customer satisfaction rose from 62% to 94% on after-hours experience surveys. Marcus also noted his staff's morale improved because they were no longer fielding angry callbacks from callers who had reached a voicemail box.
Compliance & Security
All call data is encrypted in transit (TLS) and at rest (AES-256). Call recordings are automatically purged after 90 days per PIPEDA retention requirements. Maya's infrastructure uses multi-tenant data isolation with row-level security and audit logging for every access event.
For healthcare-adjacent data (e.g., customer health info mentioned in calls), data is protected under PHIPA (Ontario) or equivalent provincial legislation. Learn more on our PIPEDA Compliance page.
Frequently Asked Questions
How much does an AI receptionist cost?
Maya starts at $399/month CAD on the Growth plan, with no setup fees or long-term contracts. This includes up to 2,000 inbound minutes per month. Overage minutes are $0.15/min. For comparison, Prairie Comfort was paying $4,500/month for one receptionist; they now pay $399/month for 24/7 availability.
Does Maya handle emergency calls?
Yes. Maya is trained to detect emergency keywords like "emergency," "urgent," or extreme conditions (e.g., "-30°C," "no heat"). When an emergency is detected, Maya prioritizes the booking and can escalate to a live technician if the customer requests it. For Prairie Comfort, emergency detection has a 94% accuracy rate.
How fast can we deploy Maya?
Deployment typically takes 2-5 business days. We integrate with your existing phone system, booking software (Cal.com), and job management tools. Prairie Comfort went live in 3 days and took their first emergency call that same evening. No on-site installation required—everything is cloud-based.
How does Mihron AI protect customer data (PIPEDA)?
All call data is encrypted with TLS in transit and AES-256 at rest. Call recordings are automatically deleted after 90 days. Mihron AI maintains audit logs for every access event. If your data includes health information (e.g., customer mentions medications), PHIPA protections apply. See our full PIPEDA Compliance documentation and the Personal Information Protection and Electronic Documents Act (PIPEDA).
Will an AI receptionist replace my staff?
No. Maya is designed to augment your team, not replace it. Prairie Comfort's receptionist still handles complex customer issues, follow-ups, and scheduling coordination. The difference: she's no longer fielding routine calls at 9 PM. Maya freed up 20 hours per week, allowing the team to focus on customer relationships and back-office work.
How do you handle payment and invoicing?
Mihron AI integrates with Stripe for monthly billing in CAD. Invoices include HST/GST. If a payment fails, Maya is automatically paused to prevent service interruption surprises. You'll receive a notification, and service resumes once payment clears.
60 Days In: What Marcus Learned Running Maya Through a Prairie Winter
After two months, Marcus had enough data to step back and assess. The numbers in the table above tell one story, but the day-to-day experience revealed a few things he hadn't anticipated going in.
The storm-surge problem got solved automatically
Winnipeg winters don't ease up gradually — they arrive in bursts. During a cold snap in late January, temperatures dropped to -38°C and Prairie Comfort received 61 inbound calls in a single 18-hour window. In previous years, that kind of surge meant Marcus and his wife were both on the phones until midnight. This time, Maya handled 54 of those calls without any staff involvement. Technicians showed up the next morning to a full dispatch board instead of a pile of callback notes to sort through.
Marcus put it plainly: "The busiest nights used to be the worst nights. Now they're just busy. Maya takes the calls, the jobs are already booked, and we show up in the morning ready to go."
Seasonal shoulder periods became revenue opportunities
HVAC businesses in Manitoba have a predictable dead zone in spring and fall — heating season is over, air conditioning season hasn't started. Marcus used the time Maya saved his front office to run a short outreach campaign for furnace tune-up appointments. Because Maya was handling routine inbound calls, his receptionist could spend four hours a day on outbound scheduling during April and May. That generated 31 tune-up bookings at $189 CAD each — work that simply wouldn't have happened without the freed-up capacity.
This pattern is common in the trades. The AI receptionist for HVAC isn't just about answering phones — it changes what your front-office staff are able to do with their time. See how this plays out across other trade verticals on our trades industry page.
What Marcus would tell other HVAC operators in Manitoba
When we asked Marcus what advice he'd pass along to other Winnipeg HVAC owners considering an AI receptionist, he had three things to say:
- Set up emergency routing before you go live. Spend an hour mapping out which call types should trigger immediate escalation. For HVAC in winter, anything involving no heat, carbon monoxide, or elderly residents in the home should go straight to a live person. That routing logic took about 45 minutes to configure and it's paid off many times over.
- Don't wait for slow season to deploy. Marcus almost waited until spring to avoid disruption. He went live in January instead and captured 23 emergency jobs in the first three weeks. Waiting would have cost him roughly $39,100 CAD in bookings.
- Check your call recordings in the first two weeks. The AI learns from context, but you'll spot edge cases early — unusual service requests, callers with heavy accents, or niche equipment questions. Flag those for your account manager and they get resolved quickly. All recordings are stored per PIPEDA guidelines and auto-deleted after 90 days.
If you're comparing AI receptionist options for your trades business, our Canadian trades AI receptionist guide walks through what to look for, what questions to ask vendors, and how to evaluate total cost of ownership in CAD — including whether a flat monthly plan or per-minute pricing makes more sense at your call volume.
Does the AI work during -40°C storm alerts when call volume spikes?
Yes. Because the system runs in the cloud rather than on local hardware, a -40°C storm in Winnipeg has no effect on call handling capacity. Prairie Comfort's highest single-day volume was 61 calls during a January cold snap — Maya answered all of them within two rings. The only limitation is your booking calendar: if every technician is already dispatched, Maya communicates realistic wait times honestly rather than over-promising. She can also collect contact details and call back when a slot opens up, which Prairie Comfort used during peak surge periods.
For context on how this compares to traditional IVR and on-call answering services, see our comparison page.
How does the AI handle emergency dispatch calls — does it know when to get a human involved?
Emergency escalation is configurable when you onboard. You define the trigger conditions: specific keywords ("no heat," "carbon monoxide," "flooding"), caller requests for a human, or situations where the AI can't confidently match the caller's issue to an available service slot. When any trigger fires, Maya transfers the call to your designated on-call number. Prairie Comfort set up two escalation tiers — tech on-call for heating emergencies, Marcus's cell for anything involving safety hazards. In three months, 23 calls were escalated to a live person; 8% of total volume. The rest booked successfully without human involvement. Pricing details for plans that include escalation routing are on our Canadian pricing page.
Ready to Transform Your After-Hours Lead Capture?
See how Maya can book emergency jobs while your team sleeps.
Request a DemoReady to get started? Contact our team for a free consultation.
Learn More
Explore how Maya works across industries: