← All work
Patient-facing Case 01 / 11 Live

A teammate, not a chatbot.

The Bilingual Patient Agent works like a real member of your front desk — almost like a person. It reads what patients actually mean, decides what to do, and books their visits over text, day or night, in English and Spanish. Not a scripted bot with canned answers — a thinking teammate that takes real action, kept honest by strict safety checks. Live with real patients since February 14, 2026.

Works 24/7
never off, never on hold
English + Spanish
switches mid-conversation
Books real visits
straight into your schedule
Safe by design
checked before every reply
The problem

A missed message is a missed surgery.

Patients text the practice at all hours, and the front desk can't keep up. At a plastic-surgery and aesthetics practice, a slow reply isn't just a missed message — it's a missed surgery. Answering texts fast was one of the biggest problems the practice had. Outside vendors had been asked for more than a year to build a texting agent, and it never materialized. So we built one that actually works.

What we built

An agent that takes real action.

Not a demo chatbot that answers FAQs — a real teammate that takes action over text. It reads what a patient means, decides what to do, and actually does it: checks the schedule, proposes times, and books the visit. It replies in English and Spanish on its own, and every message it sends is checked first, so a wrong reply gets caught before it ever reaches a patient. Under the hood, it works through each request step by step — checking, deciding, then acting — instead of jumping straight to an answer.

Replies 24/7, in two languages

Answers patients automatically in English and Spanish — including mid-conversation when patients mix both — without staff in the loop.

Books + reschedules, for real

Checks real open slots, proposes specific times, and writes the appointment straight into the practice's schedule — real actions, not just suggestions.

It thinks, then it acts

It works a request through step by step — looking things up and taking action across a few back-and-forth turns — so it doesn't have to guess at a schedule or a price.

Checked before every reply

Every outgoing text passes the same safety check before it's sent, and the wording is cleaned up on the way out — so anything off gets blocked before it reaches a patient.

Knows when to step back

If a staff member starts texting a patient directly, the agent goes quiet and lets the human run — then resumes on its own afterward.

Graded every day it runs

Every reply is graded against a fixed rubric, day after day — so any slip in quality surfaces fast instead of hiding behind a launch-day demo.

one patient text · one safe round trip
01 · Inbound
Patient texts
A message arrives, in English or Spanish.
02 · Read
Understand
Language and intent set; live patient context gathered.
03 · Act
Take action
It checks real openings, then books or reschedules the visit.
04 · Check
Safety check
The reply clears the same shared safety check, then gets cleaned up.
05 · Outbound
Reply + booking
The text goes out; the appointment is live on the schedule.
Every patient text follows the same path — read it, take action, pass one shared safety check, send. The agent can do real things on the practice's behalf because nothing reaches a patient without clearing that check first.
Sticks to approved answers Prices come from the real price list Matches the patient's language Double-checks the time slot before confirming Hands off honestly if it can't help
Why it matters
A practicing doctor put a real AI teammate in front of patients — booking visits safely, in two languages, kept live and watched closely for months.

This is the proof case for the whole idea. It went live on February 14, 2026 — confirming a real patient consult in Spanish — and it was built in a fraction of the time a paid agency quoted and never delivered. It isn't a demo: it acts on the practice's behalf, but only behind strict safety checks, and its replies are graded against a fixed rubric every day so problems surface instead of hiding. That's the exact pattern we set up for clients — customer-facing AI a business owns in-house rather than rents. The daily grading has run ever since, because trust should be re-earned every day — not claimed once at launch.

24/7
always on · English + Spanish
2
languages, one teammate
Every
reply checked before it sends
Daily
self-graded, no launch-day claim
Work with LevaIQ

Want a patient-facing agent you own?

This is the kind of system we build with practices — in your own setup, owned by you, behind safety checks that let it act without the risk. Let's find your biggest win.