AI phone receptionist for diagnostic labs
Quick summary
- At seven the chairs are already full. The person who would pick up is labelling a tube or calling the next number.
- The first question is draw, results, or a named test — those are three different jobs.
- It books the draw and takes the doctor's name if they offer it. It never interprets a result.
- Fasting is a line you typed for a named test. If you did not type it, the assistant does not invent an empty stomach.
- It checks the diary before it offers a slot, writes the time in, and texts the confirmation.
- Monday fasting and the afternoon results pile arrive together. It answers every line at once.
TL;DR
The chairs are full of people who have not eaten. You are labelling a tube. The phone rings. They want a fasting draw before work, or they want to know if yesterday's bloods are ready, or they want to know whether they may drink coffee. They will not wait on voicemail. An assistant answers in the caller's language, asks whether this is a draw, a results collection, or a named test, books a slot only if that time is free, takes the doctor's name if they offer it, and never says whether a value is normal. It does not diagnose. It does not read a result out. It does not invent a fasting rule.
Key takeaways
- Put the assistant on the published number with a forward-on-no-answer rule, so a free pair of hands at the hatch still takes the call.
- Teach it the first question: draw, results, or a named test — and do not let it jump to a time before that split.
- On a draw it takes the name, the test if they have it, and the doctor's name if offered. Give it the sentence: it cannot say a value is normal.
- Let it book bloods and a time to collect a printout. Keep interpretation and “is this okay?” as a callback from the lab or the doctor.
- After a Monday fasting burst, shorten the ring-time before forwarding so overflow is caught before the third ring-out.
- Give it the line verbatim: someone from the lab will call about results. It does not read a number, including one marked normal.
What an AI phone receptionist does in a diagnostic lab
You are labelling a tube. The next number is already standing. Across the hatch the other person is taking a history for an ultrasound. The phone rings. You cannot put the syringe down. The colleague cannot leave a patient on the chair. There is no third pair of hands.
You put the assistant on the number already on the door and on the referral slip. It answers when you cannot. Then it does one of two things. It books a draw that is actually free. Or it takes a request for results and never reads them.
The first question is whether this is a draw, a results collection, or a named test they have not booked yet. Yesterday’s printout is not a 07:30 chair. A 07:30 chair is not “is my TSH normal.” An assistant that opens with “what time would you like?” will put a results enquiry into a fasting slot and leave the person without a note.
A caller who rang in English gets the confirmation in English. A caller who rang in Greek gets it in Greek. The assistant works in 70+ languages. The SMS follows the language of the call.
Fasting morning, and the phone still ringing
One of you is already in a vein. The other is calling ticket fourteen. The published number still rings.
It is a burst. Monday after the weekend doctors’ lists. The first hour after you unlock, when everyone who must be at work by nine wants a chair. The afternoon the printers fill and people ring to ask if they may collect. A second line with the same two hands still leaves the third caller on a busy tone.
The person who wants a draw before work will not wait on voicemail. They have your number and the lab on the next street. They hang up and walk. The person whose results are already in the house will often leave a message. Those two do not cost the same.
Patients who already have a collection ticket will often wait. A person who cannot get a 07:30 chair will not. If that call waits behind the tube you are already labelling, the draw is gone, and the next lab has the tube.
Whoever takes the draw often takes the rest of that doctor’s list. That is why a missed Monday chair is not only a missed slot.
Draw, results, or a named test — then the name
Draw. They need bloods, a swab, a urine, sometimes a home draw for someone who cannot come. The assistant asks the name, the test if they have it, AMKA if you collect it on the phone, and the doctor’s name if they offer it. It checks the diary. It books a slot that is actually free. It does not say what the test is for. If you have closed new private work this morning, it says so from the line you typed in.
Results. They want to know if yesterday is ready, if they may collect a printout, if they may hear the number. The assistant takes the name and date of birth, records that they asked, and never reads a value. Ready or not ready is a status you may allow it to repeat if you marked that status speakable. What the number means is never speakable.
Named test. They have a paper from a doctor, or they remember “thyroid” and “cholesterol.” The assistant takes the words they used and the doctor’s name if offered. Preparation — fasting, water, stop the coffee — is only what you typed next to that named test. If the test is not on your list, it does not invent an empty stomach. It books the draw or flags you.
People remember the doctor’s surname. They do not always have the test’s real name. Take what they said. One identifier plus a callback is enough. None is still a note, not a guess at a panel.
A sentence the assistant is not allowed to say
“Is it normal?” is the question on this phone. It comes after a draw, after a weekend, after a doctor said “we’ll see.”
The assistant does not answer it. Not with yes. Not with “it should be fine.” Not with “TSH is usually…” A sentence that sounds like interpretation, said on the recorded line, is a sentence the lab now owns, and the assistant cannot see this patient.
The line you give it is short: someone from the lab will call. The doctor interprets. This call takes the request.
“What is this test for?”, “is a high cholesterol dangerous?”, “does this mean diabetes?”, and “can I stop the tablet?” are still clinical questions. The assistant names the topic so the callback is useful, and refuses the decision.
Do not put a reference range in the notes. It will read the range out.
Fasting is a line you typed, not a clinical opinion
Coffee. Water. Twelve hours. Six hours. “Only for glucose.” The tempting answer is a rule. The assistant does not invent one.
Kits change. Doctors write panels with names patients cannot repeat. A fasting sentence for the wrong test is a person who arrived empty and did not need to, or a person who ate and ruined the sample.
It can repeat a speakable line you attached to a named test. Unnamed work, “the doctor said bloods,” a test you do not run — those do not get a fasting lecture on this call. Take the doctor’s name if they offer it. Book the slot. Flag the preparation as unknown.
If they press — “just tell me if I can have coffee, I am in the car” — the assistant still refuses unless the line is on the list. Book them as soon as a time is free. Do not guess so they will stay on the line.
Results are never read out
Ready is not a value. Collect is not a value. “Normal” is a value wearing a coat.
The assistant may say, if you marked it, that a printout is at the hatch, or that the doctor has been sent a copy, or that the file is not ready. It may book a collection time so the afternoon hatch is not a second queue. It may not read a number. It may not say the number is fine. It may not say the number is not fine.
A patient who hears “normal” from an automated voice has been given a clinical answer by something that cannot take the next question. Configure it so that path does not exist.
If they ask whether they should go to the doctor today, that is also interpretation. Take the request. Flag a person. If they describe chest pain or trouble breathing while asking for results, that is a transfer to a phone you hold, and a line you wrote about emergency services.
The doctor’s name, if they offer it
Referring doctor is how a lab file lives. Take it when they give it. Do not demand it as a barrier to a first draw if your house does not. Do not guess Papadopoulos because that is who most of your slips say.
The assistant does not call the doctor from this conversation. It does not say what the doctor meant. It does not change a panel because the patient asked for “the full one.”
EOPYY or private is a billing split you may allow it to repeat if you typed today’s rule. It does not promise what the fund will pay. It does not invent a price for a panel you did not mark speakable.
Home draw is a diary type of its own. Distance, stairs, a morning only. Mark it. Do not let a home draw occupy a 07:30 chair in the lab.
How CITT handles it
- Your number stays on the door and on the referral slip. A forwarding rule sends what you do not answer. Calls the hatch picks up never leave the lab.
- Every line is answered at once. A Monday fasting hour is several conversations, not a stack of busy tones.
- It knows the lab you described. Which tests you run, which have a speakable preparation, whether you take EOPYY this month, who is on the chair at seven.
- It asks draw, results, or named test first, then the name, then the doctor’s name if offered, and only then looks at the diary.
- It takes results as a request, never as a number, never as “normal.”
- It checks the real diary before it offers a slot, and writes the draw into the same calendar the hatch already uses.
- It confirms by SMS, in the language of the call — 70+ languages — and can send only the preparation you marked.
- It escalates interpretation, diagnosis, “should I worry,” and chest pain as a tagged note or a transfer.
- You get the transcript, the summary, the doctor’s name if they gave one, and what the caller asked for.
What it will not do
It will not read a result. It will not say a value is normal. It will not diagnose. It will not say what a test is for. It will not invent a fasting rule. It will not tell them to stop a tablet. It will not promise what EOPYY will pay. It will not treat a draw, a results request and a printout collection as the same visit.
What these calls sound like
The order is the split, then the name, then the slot or the results note, then the refusal said out loud. There is no “it’s normal.” There is no coffee rule for an unnamed panel. There is no “don’t worry” to a person who asked for a number.
If the assistant is friendly and still answers the interpretation question, you own that sentence. Put the person who rang about yesterday’s TSH at 07:50 on a results request, not on a diagnosis.
What you set up first
Put the assistant on the number on the door with a forward-on-no-answer rule, so a free pair of hands at the hatch still takes the call. Then the three-way split, the name, and the sentence that forbids reading a result.
Write down the tests you actually run and which of them have a speakable preparation. Give it the draw fields you want on every note: name, date of birth or AMKA, test in their words, doctor’s name if offered. Mark two kinds of visit: a chair for a draw, and a short time at the hatch to collect a printout.
Give it the same-day phone list: chest pain, trouble breathing, a caller who will not stop asking for a number and is at the door. Do not give it a reference table it would read out. Write one line that forbids diagnosis, interpretation, and “normal.” Add opening hours, Monday fasting start, where to park, whether you do home draws.
Rates live on the pricing section of the home page. They are not repeated here.
AI receptionist, in-house staff, or a call centre
| CITT | In-house | Call centre | |
|---|---|---|---|
| Answers while you are labelling a tube | Always | Only between numbers | Yes |
| Several callers at once | Every line at once | One | Depends on the plan |
| Asks draw, results or named test first | Yes — before it looks at the diary | Yes | Often jumps to what time |
| Reads a result, including “normal” | Never — takes the request | The lab or the doctor calls back | Must not; sometimes does |
| Invents a fasting rule for an unnamed test | Never | Repeats the sheet if they have it | Often guesses from a catalogue |
| Speaks the caller's language | 70+ languages | One or two | Whoever is on that shift |
| What it will not do | Diagnose, interpret, or say a value is fine | Still does not diagnose on this phone | Reads a script |
What one unanswered Monday fasting hour costs a small lab
Worked from these assumptions — change any of them and the answer changes:
- Two people on reception and draw, a Monday, roughly 50 inbound calls before ten.
- About a third arrive while one of you is taking blood and the other is at the hatch.
- Of the calls nobody reaches, 6 are people who wanted a private draw this week.
- A private draw that goes ahead is worth about €35 to the lab — put your own figure in, this is the number to change first.
- This example assumes 4 of the 6 never leave a message and go to the lab on the next street.
- 6 new-draw requests reach nobody on a normal Monday fasting hour.
- 4 of them do not call back.
- 4 × €35 = €140 of that morning's private work, gone before ten.
- Monday mornings, over a 48-week year — 48 × €140.
About €6,720 a year, from the same Monday fasting window repeating.
What it costs
CITT is pay as you go: no packages, no minimum, no contract. The current rates — the monthly fee per number and the per-minute rate — are on the pricing section of the home page, which is the only place they are published.
Frequently asked questions
Do I have to give up the lab number?
No. You keep the number on the door, the referral slip, and the Google listing. You set one forwarding rule — every call, or only the ones still ringing after the seconds you choose. If someone at the hatch picks up, the call never leaves the lab and is never charged.
Can it tell a caller their result is normal?
It must not, including a result marked normal. “Normal” is the doctor's reading of a number next to this patient, not a stamp the number carries by itself. The assistant records that they asked. Someone from the lab, or the referring doctor, calls back. It does not read a value to sound helpful.
Can it say whether they must fast for a TSH?
Only if you typed a speakable line for that named test. Year, lab, and kit change what you actually tell people. A fasting sentence invented from a catalogue is a sentence you then spend the morning taking back at the chair. If the test is not on your list, it takes the doctor's name if offered and books the draw, or flags you.
What if I am already taking blood?
That is the usual morning. The assistant answers the overflow. A draw becomes a slot that is actually free. A results call becomes a note that they asked, never a number. After a Monday burst, shorten the ring-time so the third caller is not listening to a busy tone.
Someone asks what the test is for. Can it explain?
No. What a test is for is a clinical explanation. The assistant takes the name of the test as they said it, the doctor's name if they offer it, and books the draw. It does not say what the test looks for, what a high value means, or whether they should be worried.
What do we actually see afterwards?
A transcript, a short summary, whether this was a draw, results, or a named test, the doctor's name if they gave one, and whatever was booked or left for you. Confirmations go out by SMS in the language of the call.
