Building an AI Missed-Call Follow-Up Agent for Contractors
27% of home service calls go unanswered and 85% of those callers never call back. An AI missed-call agent that texts within 30 seconds recovers 1 in 3 of those jobs. Here's how to build or buy one.
Key takeaways
- 85% of callers who don't reach a live answer never call back, and less than 3% leave a voicemail
- HVAC contractors lose $3,800 a month on average from unanswered calls, totaling $45,600 a year going to competitors
- Responding within 60 seconds can lift conversion by 391% versus responding at 5 minutes, per Invoca 2025 data
When a homeowner's water heater fails at 6 PM, they dial three plumbers and hire the first one who picks up. 78% of customers hire the first contractor to respond, per Instant Sales Funnels' 2025 research. If you miss the call, you don't get a second shot.
HVAC contractors lose an average of $3,800 a month on missed calls, totaling $45,600 a year, according to Instant Sales Funnels' 2025 contractor data. The single highest-leverage AI agent you can deploy in a home service business is a missed-call follow-up agent.
Why the 30-second window matters
Invoca's 2025 home services call benchmarks show responding within 60 seconds can lift conversion by 391% versus responding at 5 minutes. After 5 minutes, the odds of qualifying a lead drop 80%, per the Dr. James Oldroyd MIT study cited in Harvard Business Review.
Less than 3% of missed-call leads leave a voicemail, and 85% never call back. Every minute your voicemail sits unanswered is a job going to a competitor.
An AI missed-call agent closes that window automatically. The customer hangs up, and a text lands on their phone inside 30 seconds. "Hey, this is Jake at Desert Plumbing. I just saw your call. What's going on, and what's the best address?"
What a missed-call follow-up agent has to do
Seven jobs in order.
- Detect the missed call from your business phone line (Twilio, Vonage, or your carrier's API)
- Enrich the caller (match to existing customer in Jobber or Housecall Pro if possible)
- Send the first text in under 30 seconds with the technician's name and an open question
- Handle the reply conversation with a real LLM, not a static chatbot script
- Extract address, issue, and urgency from the customer's replies
- Write back into the FSM as a new lead or update an existing customer record
- Hand off to a human (CSR or on-call tech) when the customer commits to a booking or the conversation hits an escalation trigger
Skip any one of those and the agent is a gimmick.
Build cost if you DIY
Build-from-scratch economics for a working missed-call agent break down roughly like this.
- 3 to 5 weeks of engineering time at $8K to $20K a month loaded, putting build cost at $6K to $25K minimum
- Twilio number and messaging at $0.0079 per SMS, typically $30 to $150 a month for a small shop
- LLM API costs of $50 to $400 a month at Anthropic's public $3/$15 per million token Sonnet pricing
- Integration work into your FSM, which usually eats 30 to 50% of total build time
- Ongoing eval work to catch regressions when the model updates or your price book changes
Ampcome's 2025 AI agent cost guide is blunt: developer time is the single biggest expense in any AI agent build, typically 35 to 40% of the total project cost.
That's before you touch TCPA compliance, which is the boring detail that kills DIY builds.
TCPA is the thing that will get you
A missed-call text-back agent is a commercial SMS message. The TCPA applies.
Quiet hours: no texts between 9 PM and 8 AM customer local time. STOP, HELP, and START keyword handling is mandatory. Express consent has to be logged and auditable. Per-violation penalties run $500 to $1,500 under the TCPA.
Every serious missed-call agent has all of that baked into the messaging layer. Your engineer cannot add it as an afterthought. Not without months of compliance work you aren't trained to audit.
Buy-side options
Podium's AI Employee responds to leads in under 1 minute and reports 45% conversion lift. Arctic Air reported a 30% revenue bump after deploying.
Hatch App's 2024 HVAC data from 132,000+ HVAC speed-to-lead campaigns shows Hatch AI CSRs reply in 5 seconds, outperforming human CSRs across the board. The best-performing campaigns hit 89.86% response rate versus an industry average of 60%.
ServiceTitan's AI Voice Agent booked 500+ jobs in 30 days for Superior Plumbing, representing 23% of their total monthly jobs without adding a CSR.
What contractors who use it say
A small residential plumber on Plumbing Zone logged 47 missed calls in a month with only 5 voicemails. After setting up an AI text-back, he recovered roughly one-third of those missed calls into booked jobs. The honest verdict: "people don't like the AI answering service, but it saves me time."
Colette Kemp's 2025 Medium piece on calling 15 HVAC companies found only 1 had professional virtual receptionist coverage. She booked with that company within minutes. The other 14 never got her business.
One HVAC owner told Smith.ai's blog that his AI receptionist "picks up every call, qualifies the job, and books it before I even know about it, with revenue up 35% since implementation."
The call-back text that actually books jobs
The prompt matters more than the model. A good missed-call text does four things in 160 characters or less.
- Uses the tech or CSR's first name, not "automated response"
- Acknowledges the missed call without apologizing for the AI
- Asks one specific open question (not "how can I help")
- Sets expectation for a human callback if the customer prefers
Good: "Hey, this is Jake at Desert Plumbing. Just saw we missed your call. What's going on at your place, and what's the address? I'll get someone dispatched."
Bad: "Thank you for calling Desert Plumbing. We missed your call. Please call us back."
The first one books. The second one goes unread.
Metrics to instrument from day one
You need visibility on five numbers or you have no idea if the agent is working.
- Missed-call detection latency (how long from hangup to first text)
- Reply rate on first text (target: above 50%)
- Conversation-to-booked-job rate (target: above 30%)
- Average booked job value through the agent (compare to inbound baseline)
- Human-escalation rate (target: under 20% but never zero)
If your vendor can't show you all five in a dashboard, they're selling you vibes.
When building still makes sense
Build in-house only if you run 10+ locations, you already employ AI engineers, and you need workflows no vendor covers. That is almost no one at $1M to $10M.
For the rest of the market, buying is the answer.
The three mistakes that kill most missed-call agents
Every failed DIY missed-call deployment hits one of these three walls.
First, generic responses. "Thanks for calling, we'll get back to you soon" has a 4% reply rate. A personalized opener with a tech name and an open question hits 55%+. The model matters less than the prompt.
Second, no FSM write-back. If the conversation happens in Twilio but never lands in Jobber or Housecall Pro, your dispatcher has no idea a job exists. The handoff breaks, the customer calls back angry, and the system is worse than voicemail.
Third, no escalation logic. Every conversation needs a clear "the AI is out of its depth, get a human" trigger. Without it, the agent keeps trying to qualify a customer who is asking a question it can't answer, and you lose trust.
The weekend-and-after-hours math
Podium's data shows most home service businesses lose 40% of leads after-hours. Your office closes at 5 PM. Customer emergencies don't.
If you're billing $4M a year and 40% of that pipeline walks away after 5 PM, that's $1.6M in captureable revenue sitting on the sidewalk every night. A missed-call agent that runs 24/7 is the floor of what every shop should be running.
How Sully handles missed calls
Sully ships a pre-built missed-call follow-up agent built for home service contractors at $1M to $10M. It connects to Jobber, Housecall Pro, ServiceTitan, Workiz, and GoHighLevel natively. It uses your real price book, service area, and customer history. It handles TCPA, quiet hours, and STOP/HELP keywords out of the box.
The agent texts in under 30 seconds. It qualifies the lead. It writes back into your FSM as a real opportunity. It hands off to a human when the customer commits to a booking.
OpenAI and Anthropic give you the language model. Sully gives you the missed-call recovery system that plugs into the tools you already run.
The verdict
A missed-call follow-up agent is the highest-ROI AI deployment in a home service business. The revenue math is brutal: you are losing $40K to $120K a year right now, and a $99 to $300 monthly tool can recover 30 to 50% of it.
Build if you have engineers and 10+ locations. Buy if you're a $1M to $10M contractor with a shared office manager and a FSM subscription.
See Sully in action if you want a missed-call agent that knows your jobs and price book the day you plug it in.
See Sully in action
Sully is the pre-built AI for home service shops. Connect your CRM, email, and phone system in minutes and the agents run on your real data.