Smith.ai vs Ruby for HVAC: Which Fits Your Shop?

Ruby states its AI boundary on its AI page: "Ruby does not use AI to… Replace human connections." Its four AI features — Call Nav (an LLM assistant that pulls answers from your business data mid-call), transcripts, sentiment analysis and robocall filtering — serve the receptionist, not the caller. Smith.ai's AI Receptionist inverts that: the AI takes the call, and "if something complex comes up, the call can be transferred to a live North America-based receptionist — any time, day or night."

Smith.ai runs a dedicated HVAC answering service page naming the intake fields it collects — system type (furnace, heat pump, AC unit), system age, service needed, urgency level, property details, preferred scheduling — the most trade-specific published detail either vendor offers. It also prints three headline percentages (82%, 50%, 67%) with no source: marketing, not evidence. Ruby's home-services page names HVAC among its supported trades but publishes no HVAC-specific intake script.

Neither vendor sells technician dispatch. Both stop in the same place: an urgency-scored call handed to someone on your side who decides who rolls. The failure mode in HVAC is not the message but the gap between message and callback, during which the homeowner keeps working down the search results.

HVAC posts a 34% phone answer rate, the second-lowest of the nine home-services trades in Invoca's 2026 benchmarks, against 52% of all inbound calls answered by a person. Invoca's figures are averages across its own customer base — businesses already paying for call analytics — but the direction holds: two of every three calls to an HVAC company in that dataset never reach a person.

Four questions that separate the two vendors on a demo call:

1. What exactly triggers "emergency"? Smith.ai's published model is a threshold you define inside a call flow; ask Ruby where urgency routing lives in your call-handling instructions.
2. Who gets woken up, and how? Smith.ai's human plans price SMS/Slack notifications at $0.50 per call, "+$0.25 for each additional 'ping' (e.g., try person B after person A)". Ruby's pricing page lists "Notifications via mobile app and email" among the features in its bundled plans, at no charge.
3. What happens at 2 a.m. on a holiday? Smith.ai prints "Calls at 2 pm and 2 am cost the same. Don't pay extra for 24/7 service." Ruby prints no surcharge for after-hours, weekends or holidays.
4. How many people can the service try before it gives up? Smith.ai's transfer destinations are metered: 1 free on Starter, 2 on Basic, 10 on Pro, $15/month each beyond. Ruby publishes flexible call forwarding with no destination count — get the number in writing.

Each bill reduces to one line of arithmetic you can run on your own phone-system report. With N calls per month at L minutes average length:

• Ruby = the smallest published minute tier covering N × L minutes. No overage rate is published, so exceeding a tier is a sales conversation, not a known cost.
• Smith.ai human = plan fee + (calls above the included count × the tier's overage rate) + $1.50 × N if you want appointments booked.
• Smith.ai AI = plan fee + (real calls above the included count × the extra-per-real-call rate). Scheduling is already included.

Ruby's published ceiling is 500 minutes — roughly 166 three-minute calls — so a 300-call month is off the price card entirely; the page says so: "Don't see the right plan here? We have larger offerings—just talk to us!"

Each vendor argues per-call versus per-minute in print, and both are partly right. Matt Lurie, a content marketer at Ruby, writes in the company's pricing-model article: "Per-call pricing tends to be misleading. This is the 'too good to be true' option. It might sound like you're paying for everything involved in a call, but in reality, you're often paying for just the call and nothing else." Smith.ai answers on each of its two price cards, in different words. The human receptionist FAQ says: "Billing per client call saves you money, and there's no mystery accounting involved." The AI Receptionist FAQ goes further: "Per-minute services work opposite to your needs; they want to maximize your call length while you want to minimize your call time."

Lurie's point holds against Smith.ai's human product specifically. Eight rows of that article's unnamed "Others" column reproduce Smith.ai's published add-on prices to the cent — $1.00 bilingual, $1.00 payment collection, $1.50 extended intake plus $0.25 per extra question, $1 web-form intake, $1.50 complex routing, $15 per transfer line, $0.50 per call per additional CRM, $5 per ad-tracking line — and a ninth prints booking at $1.50–$2.50, whose floor is Smith.ai's $1.50. Three rows fail a check against Smith.ai's own price card: Ruby's column calls the competitor "closed after-hours and holidays" (Smith.ai staffs 24/7 at no premium), gives it "3 simple questions for free" (Smith.ai's free intake covers "5 short-answer questions"), and bills it "$30 per month" plus a "$275 porting fee" for a number (Smith.ai gives a free dedicated number and will "port your business number at no charge").

Smith.ai's mirror pages are no more reliable about Ruby. Its head-to-head page — the top organic result for this comparison — lists Ruby at "$3.39-$4.90/min" and "$245/month," and its competitor table prints "Ruby $3.39 - $4.90/min." Ruby publishes no per-minute rate at all; its plan face values divide out to $5.00, $3.95, $3.60 and $3.45 per minute — a range neither figure matches, on plans starting at $250, not $245. Ruby's article contradicts itself too: the increment is "rounded up in 60-second increments" in one layout and "rounded up in 30-second increments" in another, which bill differently on a 90-second intake call.

Every comparison of these two that ranks for this query is published by one of the vendors or by a company selling a rival service — this page included. Check both price cards yourself.

Get the increment and the overage rate in writing from either vendor before you sign, then run the same math at your own volumes in how these bills behave at real HVAC call volumes.

Smith.ai's ServiceTitan integration does three things after every call, in its own words: "Create a new customer profile for callers who don't yet exist in ServiceTitan," "Match callers with existing customers who already exist in ServiceTitan," and "Add a note to the caller's customer profile with all the call information." That is customer-record hygiene, not a job on the board: nothing on the page claims job creation, capacity-aware scheduling or arrival windows, and the integration is filed under CRM — accurately.

Ruby's path is longer and named differently:

• Zapier is the spine. Ruby's integrations page says "Zapier automatically checks for new caller data in Ruby at no extra charge," and its home-services article walks the HVAC case: "Ruby takes a call about a broken AC unit," then "Zapier logs the caller in JobNimbus, schedules a site visit, and alerts your technician in Slack."
• FieldPulse is the named field-service partner, carried as a partner spotlight on Ruby's home-services page with a 15% discount for Ruby customers.
• JobNimbus is the named CRM in both the integrations list and the worked example — a roofing-and-contracting platform, not an HVAC dispatch system.
• ServiceTitan and Housecall Pro appear nowhere on the Ruby pages we fetched, including the integrations directory.

Read the integration counts as marketing. Smith.ai's HVAC page advertises "7,000+ business tools"; Ruby's pricing page advertises "5000+ integrations"; Ruby's own article credits Zapier with "8,000+ apps." All three count a Zapier-shaped universe; none tells you whether your dispatch board accepts a booked job from an outside answering service. Ask for a screen-share of a test call landing in your own instance — a five-minute check that beats any integration logo.

Segment by segment:

1. 1–3 trucks, heavy after-hours leakage, budget under $300/month. Smith.ai's AI tiers ($0 free, $150 at 75 calls, $270 at 150) are the only published option in that band. Ruby's entry plan is $250 for 50 minutes — about 16 three-minute calls.
2. Established residential shop, 100–200 calls a month, brand voice matters. This is Ruby's strongest case: one bill, scheduling and bilingual answering and payment collection included, no add-on menu. Expect $720 to $1,725 a month; get the overage rate in writing, since Ruby publishes none.
3. Commercial or multi-stakeholder work. Smith.ai's human-first plans, with escalation thresholds you define, are the closer published fit — its HVAC page hands live receptionists "commercial system installations involving multiple stakeholders, or warranty and billing discussions that require careful handling."
4. 300+ calls a month. Neither vendor's published card covers you. Both become custom quotes; Smith.ai's AI Enterprise steps ($500 at 300 calls, $800 at 500) are at least printed.
5. You need the job on the board, not a note in the CRM. Neither fits; that capability sits with the trade-specific AI vendors. We ranked the eight-vendor answering-service field we tested, and where those agents break on an HVAC line is in what an AI phone agent can and cannot close out.

Where we fit: we run a single-trade answering service, HVAC only — a competitor to both companies above, which disqualifies us as the neutral party. We publish no performance guarantee and we do not claim a booking-rate figure we cannot show you. Unlike the two vendors above, we print no price card for it either: our AI answering work is quoted per shop, because call volume and after-hours load differ too much between a two-truck shop and a twenty-truck one for a published tier to mean anything. Send us your monthly call count through the contact form and the quote comes back with its assumptions written out. Our agent books into ServiceTitan and Housecall Pro, and only those two — no other field-service platform is wired, and we hold no certified-app badge on either.

Put your own call volume and average ticket into the missed-call math for your own shop first. If unanswered calls cost you less than $250 a month, neither vendor's human plans pay for themselves yet — start on Smith.ai's free AI tier and re-run it in a quarter.