Automation that answers the phone
AI Lead Intake & Automation Pipelines
Productized AI and automation systems for local service businesses: a live conversation-AI intake bot, missed-call text-back, review funnels, and CRM pipelines — engineered on GoHighLevel and measured in booked jobs, not demos.
- Context
- Founder / Operator
- Timeframe
- Ongoing
- GoHighLevel
- Conversation AI
- Webhooks / API integrations
- Twilio-style SMS (A2P)
- CRM pipeline design
- Node.js scripting
The setting
For a local service business, the most expensive software failure isn’t a crash — it’s a missed call. Somebody’s water heater dies, they call three plumbers, and the one who answers first wins. Most of my automation work lives right at that moment.
What I build
“Alex,” the AI intake bot. Built on GoHighLevel’s Conversation AI and live in production for McCool’s Pool Service: it answers inbound leads, asks the qualifying questions a good office manager would, and files the result into a CRM pipeline with the follow-up already scheduled. It works the hours when no human can pick up — which for a field-service company is most of them.
The automation layer around it. Missed-call text-back (the caller instantly gets a text before they dial the next company), review funnels that turn finished jobs into Google reviews, and full pipeline builds from first contact to booked job. I run this as a GHL reseller operation, which means the unglamorous engineering too: sub-account provisioning, A2P phone-number registration and compliance, webhook wiring, and the API integrations that make each client’s stack fit their actual workflow.
Retention loop
The engineering mindset behind it
Two things separate this from plugging in a chatbot:
Unit economics as a design input. Every AI conversation has an API cost, and a productized service has a fixed monthly price — so cost per conversation is a margin variable I model before I build. Prompt design, conversation length limits, and escalation-to-human rules are all partly economic decisions. I treat “does this pay for itself” as an engineering requirement.
Untrusted input, always. An intake bot reads whatever the public sends it. Anything a lead types is data, never instructions — the same prompt-injection discipline I apply in my agent orchestrator applies at the front desk.
Why this project matters to an employer
This is applied AI with a measurable job: revenue captured that used to evaporate. It shows I can take a new capability, wrap it in the compliance and integration work nobody tweets about, price it, and run it in production for businesses that notice immediately when it breaks.