The front desk is short-staffed. What an AI receptionist can and cannot do

The ADA Health Policy Institute's economic outlook polling carries one number every practice owner recognises: about one in three practices report actively recruiting front-office staff amid a persistent shortage. The front desk is where the shortage shows first, because the phone is where patients decide whether to come in. The 411 Locals study of small businesses found 62% of calls not answered by a person. A practice with one receptionist, a patient at the window and two lines ringing is not an exception to that figure.

What the phone actually does in a practice

Most calls are not emergencies. They are scheduling and rescheduling, insurance questions, price questions, recall reminders answered late, and confirmations. Each takes a few minutes from a person who is also checking patients in and out. When the volume exceeds the person, calls roll to voicemail, and the HBR audit of 2,241 companies suggests what happens next: the average response to a new inquiry was 42 hours, and a prospective patient has usually booked elsewhere by then.

What an AI receptionist can do

Answer every call on the first ring, including after hours. Book, move and cancel appointments directly in the practice management system, within the rules you set for chair time and provider. Answer the routine questions from your own knowledge base: hours, location, the insurance networks you participate in, what to expect at a first visit. Run recall and confirmation calls and texts so the hygiene schedule stays full. Log every conversation to the patient record. Hand the call to a person, immediately, when a caller asks for one or when your rules say so.

What it cannot and should not do

It cannot give clinical advice and it should not try. A caller describing pain gets a same-day slot and a human, not a diagnosis. It cannot resolve a complex insurance dispute; it can collect the details and route them. It cannot replace the judgment of a good front-desk lead who knows which patient needs ten extra minutes. Anyone selling you a receptionist that handles everything is selling you a liability.

Design the handoff before you deploy

The useful question is not whether AI can answer the phone. It is which calls should reach a person, how fast, and with what information already collected. Write those rules first: emergencies, upset patients, anything involving a treatment plan. The engine follows them. Your team gets the calls that need a human, with the routine ones already handled.

Measure it like a hire

Baseline a month: calls, answer rate, bookings from the phone, no-shows, and the hours the front desk spends on the phone. Deploy, then compare. A receptionist you hired would be judged on the same numbers.