Voice AI & automation

I build phone agents that answer the call, finish the task, and hand off cleanly.

Voice receptionists for hotels, clinics and service businesses, built on Retell, n8n and GoHighLevel. I scope the job with the client, build it, and run the go-live call.

Based in the Philippines, working US hours.

  • Retell
  • n8n
  • GoHighLevel
  • Zapier
Portrait of Jet Corsiga

Demo walkthroughs · coming soon

Watch my demo agents

Short recordings of real test calls, followed by the flow and the n8n workflow behind each one. Nora, the hotel receptionist, turns down a same-night booking after 9 PM. Maya, the clinic receptionist, books an appointment and checks times again when the caller changes the day. Same kind of flows I build for clients.

Fictional businesses · no real data

Selected work

Client names are removed. Demo builds use fictional businesses on my own accounts.

Live Client work Retell Conversation Flow · n8n · GHL

After-hours and overflow AI receptionist for a boutique hotel

Problem. Calls the front desk couldn't pick up, overnight and at busy times, went unanswered. Booking requests and guest issues were lost, and a booking-system API would have added a monthly fee.

Built. An agent that checks at call start whether the desk is open and changes what it promises. Six paths: questions, new bookings, changes, cancellations, callbacks and urgent issues. Each request is read back, then n8n emails it to staff. Callers are verified before any reservation request, and the agent never pretends it can book live.

Result. Live and answering real guest calls after hours and when the desk is busy. Requests reach staff by email instead of being lost.

Live Client work Retell Conversation Flow · n8n · CRM calendar

Inbound receptionist with live booking for a chiropractic clinic

Problem. The clinic wanted callers to book an appointment, leave a message or reach the office without waiting on the front desk.

Built. A 20-node flow that books appointments during the call: find or create the patient, check availability for the exact date, offer times, confirm, book. The appointment type is picked from visit history. Separate paths handle messages, office callers and live transfer, with fallbacks when the office is closed.

Result. Live and taking real patient calls, booking appointments straight onto the clinic calendar during the call.

In build Demo build Retell Conversation Flow · n8n

Driftwood Cove Inn: collect-and-forward receptionist

What it shows. A fictional 24-room coastal inn. The agent, Nora, takes after-hours calls and daytime overflow. She answers questions from a knowledge base and logs bookings, changes, cancellations and callbacks.

Design choices. Every logging path reads details back before saving. n8n falls back to caller ID when the caller gives a partial number. A same-night request after 9 PM gets a clear no, never a false booking.

Evidence. Video walkthrough coming soon.

In build Demo build Retell · n8n · Google Calendar

Cedar Ridge Family Chiropractic: live-booking receptionist

What it shows. A fictional clinic. The agent, Maya, checks availability on a calendar and books new-patient exams and adjustments during the call. Reschedules and cancellations are taken as messages for staff.

Design choices. n8n does all date and time conversion and returns ready-to-read strings, so the agent never does time math. If the caller changes the date, availability is checked again. A failed booking sends an alert so a human follows up.

Evidence. Video walkthrough coming soon.

How I build

Rules I follow because each one fixed a real bug.

  1. The agent never does time math

    Dates, time zones and slot formatting live in n8n. The agent reads back exact strings it was given.

  2. New date, new check

    When a caller changes their mind, availability is looked up again. Old slots are never reused.

  3. Confirm before logging

    Collect, read back, then log. If the caller corrects a detail, the agent confirms again before anything is sent.

  4. Plan for blank fields

    Callers give partial numbers. The workflow falls back to caller ID instead of failing or dropping the request.

  5. Never promise what isn't wired up

    If the agent can't book live, it says it's passing the request to staff. I scope that honestly with the client up front.

  6. Failures reach a human

    When a booking or webhook fails, the caller gets a polite fallback and staff get an alert.

Working with me

In my full-time role I join onboarding calls as the technical lead, work out what is possible and what setup the client needs, build the account, and run the go-live call. I send a written end-of-day report every shift, so you always know where things stand.

Have a phone line that needs an agent?

Tell me about the business and what happens when the phone rings today. I'll tell you what can be built and how.

Intro call: Voice AI project. 30 minutes on Google Meet. Slots are Tuesday to Saturday, 6:30 to 9:30 AM Philippine time, which is Monday to Friday afternoon and evening in the US. The booking page shows times in your own time zone. Button not opening? Open the booking page in a new tab.