The problem
The practice runs several locations with a shared front desk. Clinicians are in session most of the day, so the phone competes with patient care. Callers wanted appointments, prescription refills, billing answers, referrals and simple messages passed on, in English and in Spanish. Each one needed a person to stop what they were doing, and after 6 pm nobody was there at all.
Most calls arrive between noon and 8 pm, while patients are in the building and the desk is already busy.
Refill, billing and paperwork requests took the same staff time as a new patient, with none of the value.
Spanish-speaking callers depended on whoever happened to be free.
Messages were taken on paper or in email and reached the wrong person, or nobody.
After-hours callers reached voicemail, and many did not call back.
What we built
We treated the front desk as a set of workflows rather than one phone line. Each kind of call, from a booking request to a forensic referral, was mapped to who handles it and what the agent is allowed to do on its own. The agent greets callers in English or Spanish, works out what they need, and either finishes the task itself, writes a structured note for a named staff member, or transfers the call live. Every call is recorded and summarised, and the practice sees the whole picture on one dashboard.
Inbound call
Practice line, bilingual
AI agent
Caller
AI agent
Caller
AI agent
Caller
AI agent
What the system did
- Spanish detected, agent switched language
- Intent: booking, existing patient
- Two open slots found in the calendar
- Appointment created
- Summary and recording saved
Receptionist agent
Answers every inbound call around the clock, in English and Spanish, with a professional tone the practice chose.
Appointment booking
Checks the practice calendar and books, reschedules or cancels during the call.
Intent workflows
Separate paths for new patient intake, existing patients, refills, billing, administrative requests and referrals, each routed to the staff member who owns it.
Message taking
Structured notes with caller name, number and request written to a shared worksheet the team works from.
Live transfer
When a caller asks for a person, the call goes to the named staff member with the context attached.
Call intelligence
Recordings, summaries, sentiment signals and an hourly volume view, so the practice can see peaks and adjust staffing.
Spam and emergency handling
Spam calls filtered; emergency language triggers an immediate handover.
How it works
From trigger to result.
How the system works
7 stages
01Inbound call
ChannelRings the practice line at any hour, in English or Spanish.
02Language and intent
AIThe agent greets the caller and works out what they need.
03Booking
LogicOpen slots from the practice calendar, booked during the call.
04Request routing
LogicRefill, billing, intake and referral requests go to the staff member who owns them.
05Team worksheet
DataStructured note: who called, from which number, what they need.
06Live transfer
PersonA caller who asks for a person reaches them, context attached.
07Dashboard
ResultRecording, summary, sentiment and hourly volume for every call.
Our approach
How we worked.
Discovery
Listened to how calls actually arrive, which requests repeat, and who on the team owns each one.
Workflow design
Mapped every call type to an action the agent may complete, a note it must write, or a person it must reach.
Wording and knowledge
Approved greetings, answers and limits in English and Spanish.
Soft launch
Live on the main line with recordings reviewed daily and workflows adjusted.
Steady state
Monthly review of the dashboard and the agent's own improvement suggestions.
The outcome
What changed.
Every call is answered, in English or Spanish, at any hour, including the 56 that arrived after the desk closed.
Routine requests reach the right staff member as structured notes instead of voicemails.
Callers who need a person reach one: 285 transfers in the month, each with the context the agent gathered.
The practice sees volume by hour and by intent, so staffing follows demand rather than guesswork.
Same receptionist, different business
The pattern does not depend on the industry. For a senior move management company we built the same kind of agent on their own CRM, answering after hours and booking consultations live.
Home & Local ServicesAI phone and chat agent that captures leads after hours
Questions
What people ask.
Can an AI receptionist handle calls that involve health information?
Yes, with the right platform. For clinics and practices we deploy on infrastructure that meets HIPAA, SOC 2 Type 2 and PIPEDA standards, with call recording under the practice's control and access limited to named staff. We review the setup with the practice's own compliance adviser before going live.
What happens when a caller needs a person?
The agent transfers the call to the staff member responsible for that kind of request, with the context it has gathered. If that person is unavailable, it takes a structured message so nothing is lost. In this practice, 285 of 759 calls in a month were transfers.
Does it work in more than one language?
This agent answers in English and Spanish and switches to the caller's language. Other languages can be enabled when the practice needs them.
How does the practice know what the agent said?
Every call has a recording, a written summary, the intent it detected and a sentiment signal. The team reviews these from one dashboard and can flag any call for follow-up.
