How to Train a New Dental Receptionist Faster With a Chatbot Handling the Repeat Questions

It's Jen's first Monday at Smile & Skin Clinic. At 9:40 a message comes in on the clinic's Facebook page:

"hm po pa cleaning?"

She doesn't know. The receptionist who left last month kept the prices in her head. So Jen knocks on the treatment room door and asks Doc, who has gloves on and a patient in the chair.

At 10:15 it's "open po kayo ng Sunday?" At 10:30, "accept po kayo Maxicare?" At 11:00 somebody wants to know if there's parking. By lunch Jen has interrupted the dentist six times, nine threads are still waiting for a reply, and she's quietly wondering whether she took the right job.

None of those questions were hard. Jen isn't slow because she's new to reception work. She's slow because she's new to your clinic, and nobody ever wrote your clinic down.

The short answer

Stop making a new receptionist learn the repeat questions by interrupting people. In practice that's five steps:

1. Count the questions patients ask most, using your own inbox.
2. Write one answer sheet with every repeat question on it, answered the way you want it answered.
3. Have a chatbot answer those questions from that sheet, and pass everything else to a person.
4. Have your new receptionist learn the sheet by reading the chatbot's conversations each morning.
5. Spend her training time on the work only a person can do.

The sheet does two jobs. The chatbot answers from it, and your new hire learns from it. You only write it once.

Why a new receptionist is slow in week one

A front desk runs on three kinds of knowledge, and each one takes a different amount of time to learn.

Kind	Examples at Smile & Skin Clinic	How long it takes to learn	Can a chatbot answer it?
Facts	Cleaning price, clinic hours, which HMOs you accept, where to park, what to bring	Days, if they're written down	Yes, from an approved list
Rules	A cleaning takes 45 minutes. Braces adjustments only on the orthodontist's days. No extractions after 4 PM	Weeks, mostly by getting corrected	A booking system can enforce most of them
Judgment	A patient in pain at 5:55 PM. A complaint about last week's bill. A parent whose child cried through the last visit. An HMO approval that needs a phone call	Months, by watching someone good do it	No

Most first messages to a clinic are about facts: a price, a schedule, a location. You'll check that against your own inbox in step 1. What makes a receptionist good, though, is judgment, and learning on the job gets the two backwards. The facts arrive all day, one interruption at a time, and they eat the hours she should be spending watching how the judgment calls get made.

Facts are also where a new hire's mistakes are most public. A wrong price said at the counter gets corrected on the spot. A wrong price typed on Messenger stays in writing, and patients screenshot.

Step 1: Count what patients actually ask

Don't write the list from memory. Owners tend to remember the unusual questions, and the common ones blur together.

Open your page inbox and go back 30 days. For each thread, jot down the patient's first question in a few words, then tally them. If patients also text the clinic phone or message you on Viber, go through those too.

At Smile & Skin Clinic the top of the list would look something like this:

• "hm po pa cleaning?" and "pila ang pa-cleaning?"
• "open po kayo ng Sunday?"
• "may slot po ba bukas?"
• "accept po kayo Maxicare?"
• "saan po kayo located?"
• "magkano po braces? may installment?"
• "pwede po walk-in?"

The count gives you two things. The first is the list itself: the questions that come up again and again, leaving out the one-offs. The second is how patients word their questions, which is rarely how you'd word them. Nobody types "What is the fee for oral prophylaxis?" They type "hm po pa cleaning". Keep their wording, because you'll use it in step 2.

If the receptionist who left worked from memory, this is also how you recover what she knew. It's all in the thread history: every price she quoted, every "sorry po, closed kami ng Sunday", every time she told someone the dentist was on leave. Copy the answers out, then check each one is still true before it goes on the sheet.

Step 2: Write the answer sheet once

Put every repeat question in one document, with the answer you want given. A shared Google Doc works. So does a printed page in a folder at the desk, as long as there's only one and everyone works from it.

Give each question four columns:

Column	What goes in it	Example
The question, as patients type it	Their wording, including Taglish and Bisaya	"hm po pa cleaning", "pila ang pa-cleaning"
The answer	The full reply, in the clinic's voice	"Cleaning is ₱1,500 po and takes about 45 minutes. If there's heavy tartar, Doc may suggest a deep cleaning and will quote it after the check-up."
Who owns it	The one person who decides when it changes	Doc
Last checked	A date	1 September 2026

The owner column is the one people skip, and it's the one that matters six months in. When the cleaning price goes up, somebody has to change the sheet that same day. If nobody owns the answer, nobody changes it, and your receptionist and your chatbot both keep quoting last year's price.

Write each answer out in full, the way you'd want it said. "Mon-Sat 9-6" is a note. "Open po kami Monday to Saturday, 9 AM to 6 PM. Closed Sundays and holidays." is an answer. The note tells a new receptionist the hours. The answer also shows her how the clinic talks, and it handles the holiday question before anyone asks it.

Then write a second, shorter list: the messages that never get a written answer and always go to a person.

• Pain, swelling, bleeding, or any "normal po ba ito?"
• Photos of teeth or gums
• "Is this safe for me?" questions about a treatment or a medicine
• Complaints and refund requests
• "Pwede po discount?"
• Anyone who asks for a person

That list teaches her as much as the first one, because it shows her where the chatbot stops and her job starts. Our walkthrough of how an AI receptionist works has a longer version of it, and what to answer at 10 PM and what to hold until morning covers the after-hours side.

Step 3: Let a chatbot answer from the sheet

Switch the chatbot on before your new receptionist's first day if you can, so she doesn't start with a backlog.

The word that matters is "from". A chatbot that answers only from your sheet quotes the ₱1,500 you wrote. A chatbot that answers from general knowledge can come up with a plausible number instead, and your new receptionist can't catch it, because she doesn't know the real number yet either. That's how a clinic ends up honouring a price nobody set. If you're comparing tools, our guide to choosing a Messenger chatbot for a dental clinic lists the test messages to send before you sign anything.

Set one handoff rule from the start: anything that isn't on the sheet goes to a person, with the patient's name and number taken. Then pay attention to those handoffs. Each one is a gap in the sheet, and it's also a question your new receptionist would have had to ask somebody. Fill the gap once and the chatbot and the receptionist both know the answer from then on.

For Jen, the difference shows up on day one. The 10 PM "hm po pa cleaning?" is answered before she arrives. The 9:40 AM one is answered while she's checking a patient in at the counter. She's no longer the bottleneck for questions she can't answer yet.

If you want to see this from the patient's side, the chat on the OtterFlow home page is live, running on our Smile & Skin Clinic demo. Ask it a price, then ask it something it was never told.

Step 4: Train her on the conversations, not the interruptions

The obvious worry comes next. If the chatbot handles the repeat questions, how does your new receptionist learn them?

She reads. For her first two weeks, she spends fifteen minutes every morning going through yesterday's conversations: what patients asked, what the chatbot said, and which threads got handed over and why. It's the same material she'd have picked up by knocking on Doc's door, except it's in order and written down, and it doesn't cost anyone a clinical hour.

From the second week, turn it around. Pick a handful of threads, have her give the answer she'd have sent, and check it against the sheet. Once she can do that without looking, she knows the facts. With a written sheet that takes days. Without one, it takes as long as it takes her to hear every question at least once.

She should still know every answer on the sheet by heart. The chatbot covers Messenger at night and while she's busy. It doesn't cover the patient at the counter asking the same price, or the one who rings the landline.

Teach her takeovers in the same week. When a thread needs a person, she should take it over properly, so the chatbot stops replying in that conversation until she hands it back. Without that, you get two voices in one thread: Jen typing a reply while the chatbot is still answering underneath her. To the patient, that looks broken.

Step 5: Spend the saved hours on judgment

The time you get back belongs to the third row of that table, the part no document covers well.

• The counter. Checking patients in, taking payment, and asking about the next visit before they leave. That question is the cheapest way to get a second visit booked, and we cover it in turning first-time patients into returning ones.
• The phone. HMO approvals especially, since those often mean a call to the HMO while the patient waits.
• The handoffs. A toothache message at 8 PM never gets advice from the front desk. It gets a person. She needs to know who that person is after hours and how fast they should hear about it.
• Money and complaints. Refunds, a disputed bill, a patient asking for a discount. Tell her plainly which of these she can decide and which come to you.
• The calendar's rules. Which dentist does which procedure, how long each one takes, and what can't go back to back. A booking system can enforce these, but she still has to understand them when she moves an appointment by hand.

The best way to teach this part is still the old way. She watches you or a senior staff member handle a few, then handles a few while you watch, and you go over them together at the end of the day. It's slow, and it should be. This is the part of the job worth training slowly, and it's the part the repeat questions used to crowd out.

A first month that works

When	What she focuses on	What the chatbot handles
Before day 1	Nothing yet. You count the questions, write the sheet and switch the chatbot on	Repeat questions, from the sheet
Days 1 to 3	Reads the sheet. Reads yesterday's threads each morning. Shadows at the counter	Repeat questions on Messenger, day and night
Week 1	Checks patients in. Handles handed-over threads with help	Same. Each handoff shows a gap to fill in the sheet
Week 2	Answers from the sheet without looking. Starts taking over threads that need a person	Same, with fewer gaps
Weeks 3 and 4	Phone calls, HMO approvals, moving appointments by hand, first complaints with you nearby	Same
Months 2 and 3	Handles handoffs on her own. Suggests changes to the sheet	Same

If you're hiring on probation

Philippine law caps probationary employment at six months. The implementing rules also say the standards she has to meet to become a regular employee must be made known to her when she's hired. If they weren't, she's deemed a regular employee. Labor Law PH's guide to probationary contracts walks through Article 296 of the Labor Code and the rule behind it.

The sheet and the handoff list give you something concrete to put in writing on day one. For example: by the end of month three, she answers every question on the sheet without looking it up, and she knows which messages go to a person. Standards are only one part of a probationary contract, so have whoever handles your HR paperwork check the rest.

Common mistakes

• Training by interruption. Every "Doc, magkano po ulit yung...?" costs a clinical minute and teaches one fact to one person.
• Letting the chatbot answer from general knowledge. Your new receptionist can't catch its mistakes, because she doesn't know the right answers yet either.
• Writing the sheet in the dentist's words. "Oral prophylaxis" is what's on the price list. "Pa-cleaning" is what patients type.
• Leaving answers without an owner. The price changes, the sheet doesn't, and the chatbot and the new hire both quote last year's figure.
• Two voices in one thread. Decide who's answering each conversation, and hand it over properly when that changes.
• Skipping the facts because "the chatbot knows them". She'll still get asked the price at the counter, on the phone, and at the mall by a neighbour who knows where she works.

Where OtterFlow fits

When we set OtterFlow up for a clinic, we build the knowledge base from your FAQs, which is the answer sheet from step 2, and we train your staff as part of setup. Flo, our AI front desk, answers patients on your web chat or Messenger page from what you approved, in English, Taglish or Bisaya, matching however the patient writes. It books into times that are genuinely open and sends the SMS confirmation and reminders. When it doesn't know something, it says so, takes the patient's details and flags the thread for your team.

Your new receptionist can take over any conversation, and Flo stays quiet in that thread until she hands it back. That covers the handoffs in step 3 and the takeovers in step 4.

Setup: ₱0. Then 4,975 a month.

Questions owners ask

Does this work without a chatbot?

Partly. The answer sheet on its own speeds up training, because she reads instead of asking. What you lose is cover while she learns. The 10 PM messages still wait until morning, and the 9:40 AM ones still wait while she's busy at the counter.

What if the chatbot gives a wrong answer?

Find out where it came from. If it quoted the wrong price because the sheet had the wrong price, fix the sheet, and the chatbot and your receptionist are both right from then on. If it gave an answer that isn't on the sheet at all, that's a design problem, and it'll happen again. Test for it before you commit: ask the chatbot something you never told it and see whether it admits it doesn't know.

Will a new receptionist feel replaced?

Tell her on day one exactly what the chatbot does. It takes the questions she can't answer yet. It doesn't take the counter, the phone, the patients who need a person, or the judgment calls, and those are what you hired her for.

The bottom line

A new receptionist is slow because your clinic lives in somebody's head. Count the questions, write the sheet before she starts, let a chatbot answer from it, and spend her first month on the parts of the job a page can't teach.

If you'd like help writing that sheet, and having it answer patients by the time your new hire starts, book a 15-minute call and we'll go through what your patients ask most.