A bilingual voice agent that books the class
The same Barcelona Pilates studio wanted its phone line answered like its DMs. We built Sophie, an inbound voice agent on the studio's Spanish number: she explains the classes, collects and spells back the caller's details, checks the live calendar, offers another slot if the first is taken, and books the free assessment.
From hello to booked, in about two minutes
The prompt is built around one goal, a booked free assessment, with a fixed order so the agent never checks the calendar before it has a name and an email it trusts.
One prompt, one tool, one goal
A single-prompt agent on GPT-4.1 with a native Spanish voice. One custom function does the real work: it posts the requested time to a Make webhook, which checks and books the GoHighLevel calendar and returns open slots.
Spell it back
Emails are read back letter by letter before the domain, the most common place voice bookings go wrong.
Budget and time handled
The free assessment answers price worries; evening and weekend slots answer schedule ones.
Fails out loud, politely
If the calendar check fails, she says so and rechecks the details instead of guessing.
Two scenarios behind the voice
During the call, checkAvailability hits the booking scenario: OpenAI turns the spoken time into a timestamp, the caller is found or created in GoHighLevel, and the slot is booked or two alternatives come back. When the call ends, a second scenario logs it to Airtable and updates the contact.
Slot taken. Booked anyway.
A call on the live line: the first time asked for was taken, so she offered two others, the caller picked a different day, and the booking went through in 2 minutes 15 seconds. After every call Retell extracts the caller's name, email, reason for calling, chosen time and whether a booking was made.
Every call ends as data
Live on the studio's Spanish phone line. Callers get answers and a booked assessment without waiting for staff, and the studio sees outcomes, not just recordings.
Retell · GPT-4.1 · Make · Airtable · GoHighLevel
Custom function, post-call extraction and logging, calendar booking and confirmation email.
Confirm before you book
The calendar is only queried once the name and email are confirmed.
Same studio as the chat agent
Built alongside the DM agent in case study 05, so phone and chat book the same way.
Three things we learned
The full case study as a PDF
The full write-up with the screenshots above, formatted for printing or forwarding. Client details are hidden in the PDF too.
Download the PDFMissing calls, or answering the same booking questions all day?
Tell us what you are trying to automate or move. We will say what fits, what does not, and what it would cost.
Book a free 15-minute call