AI for Urgent Care: Wait-Time Texts, Check-In and Staff Hand-Offs

An urgent care clinic can use AI for wait-time texts and check-in by keeping it to patient-flow messages only: you are in the queue, here is the estimated wait, we are ready for you. It must not triage symptoms. A short list of red-flag words sends any reply straight to a named staff member and tells the patient to call the emergency number.

That boundary is the whole design. Patient flow is a logistics job: counting, timing and telling people. Triage is a clinical job: deciding how sick someone is. A system that blurs the two, even politely, creates a clinical safety problem. The steps below build the logistics side and fence the clinical side off.

Follow me on Instagram@sagnikteaches

Stage 1: decide what the messages are allowed to say

Write the list of message types before you pick a tool. A typical walk-in clinic needs only six, and each is a template a clinician approves once.

Connect on LinkedInSagnik Bhattacharya

1. Check-in confirmation. "You are checked in at [clinic]. Your place in the queue is [number]. Estimated wait [range]. Reply STOP to opt out."
2. Wait update. Sent when the estimate moves by more than a set amount, say 20 minutes.
3. Almost your turn. "Please head to reception now. We will call you in about [x] minutes."
4. Your turn. "We are ready for you. Please come to the front desk."
5. Missed call. "We called your name. Please see reception within [x] minutes to keep your place."
6. After the visit. Opening hours, a number for administrative questions, and a reminder that any worsening problem needs urgent attention.

Notice what is missing: any sentence about symptoms, any advice, any "we hope you feel better". Those drift into clinical territory and tone problems. Short messages are also cheaper: a text over 160 characters is split and billed in segments, and a single curly apostrophe or emoji can drop the limit to 70 characters per segment (67 once a message is split). Twilio's optional Smart Encoding swaps curly quotes for straight ones, but writing in plain characters from the start keeps each message to one segment whatever the provider.

Subscribe on YouTube@sagnikteaches

Stage 2: separate the walk-in queue from the online queue

Most clinics run two queues that behave differently. Walk-ins arrive and wait. Online check-in patients hold a place from home and arrive later. If your messages treat both the same, someone ends up waiting an hour while the screen says ten minutes.

Decide the policy first: do online check-ins join a single queue in the order they arrive, or do they get time slots? Either works, but the text must match it. Say a clinic blends the two, treating a check-in at 10:05 like a walk-in at 10:05. Then the message at 10:05 reads "your place is number 7" and a sixteen-minute drive to the clinic might lose the patient three places. Fix this by sending "leave now" at the moment the patient's position matches the travel time they entered.

Stage 3: pick the tool, and what AI actually does in it

Many urgent care clinics already hold a platform with online check-in and wait-time display built in. Platforms sold to walk-in clinics include Solv, whose provider pages list a digital check-in kiosk, a mobile waitlist and patient messaging, and Experity, whose patient-engagement software offers online check-in and wait-time updates (the older Clockwise.MD check-in site now redirects to Experity), among others; confirm what yours offers, and whether texts are included or billed separately, on its own pricing page or with your account manager. If you hold one, use its templates before buying anything else.

If you do not, a small clinic can send queue texts with a general messaging service such as Twilio, fed from a simple sheet or your practice system, though that needs someone comfortable connecting the pieces. Check the provider's rules on business texting and number registration with them before you start, because they differ by place and affect delivery.

AI's role in this setup is small and well defined:

• Writing and tightening templates in plain characters, at a reading level patients follow on a phone.
• Drafting the staff hand-off notes in a consistent shape.
• Classifying replies narrowly: is this a logistics reply (running late, I have left, I cannot come) or does it contain a red-flag word? It never decides what a symptom means.
• Summarising the day's queue data for the manager: peak hours, longest waits, no-show rates.

It does not forecast how sick a patient is, set priority in the queue or give advice. For the numbers side, see the wait-time dashboard tutorial and volume forecasting for rotas.

Stage 4: build an honest wait estimate

A wait estimate is an average dressed up as a promise, so build it simply and label it. The most common rule is: patients ahead of you times the average minutes per patient for this hour of the day, shown as a range.

A worked example with invented numbers. A clinic sees an average of 5.5 patients an hour at lunchtime with one clinician and a nurse practitioner working. That is about 11 minutes per patient on average. A patient checks in with six people ahead. The estimate is 6 times 11, or 66 minutes. The text says "about 55 to 80 minutes", a range of roughly a sixth below to a fifth above, because a range tells the patient it is not a guarantee. If a chest-pain patient arrives and takes the clinician for 40 minutes, the next update goes out when the range moves by 20 minutes: "Your estimated wait has changed to about 80 to 100 minutes. We are sorry for the delay."

Two checks keep the estimate honest. Compare the sent estimate with the actual wait for each patient for two weeks, and adjust the minutes-per-patient figure by hour. And never display a lower number to attract walk-ins than the one you tell checked-in patients. The first time a patient notices, trust goes. The numbers feed neatly into the dashboard described in the KPI tutorial linked above.

Stage 5: the red-flag rule that keeps clinical questions with staff

Patients reply to texts with whatever is on their mind: "my chest is getting tighter", "my son is not breathing right", "still bleeding". A logistics system that answers "thanks, your wait is 40 minutes" is a safety failure. The fix is a rule, owned by a clinician, that runs before any other logic.

1. Write the list. A clinician writes the red-flag words and phrases for your clinic: chest pain, trouble breathing, cannot breathe, bleeding, fainting, severe allergic reaction, suicidal, overdose, stroke signs, head injury and others appropriate to your service. Err on the side of catching too much.
2. Catch spelling variation. Texts are misspelled and abbreviated. Include common forms ("cant breath", "chest hurts") and have the clinician review the list each quarter against real replies.
3. Send two things at once. To the patient: an automatic reply such as "If this is an emergency, call the emergency number now. A member of our team will contact you shortly." To staff: an immediate alert to a named person, by phone or pager, not an inbox.
4. Stop the automation. After a match, no further automatic queue messages go to that patient until a human clears it.
5. Test it monthly. Send test messages containing each phrase and check how fast a human responds. Log the result.

A word-list will miss things, and AI classification will miss different things. That is why the rule is two layers: a fixed list plus, if you wish, an AI check that errs on the side of flagging when a message mentions the body, pain or a child in distress. When in doubt, the message goes to a person. A false alarm costs a minute of staff time. A missed one is the incident you are designing against.

Stage 6: the staff hand-off note

Queue texts are one half of the job. The other is what staff see. A good hand-off view tells the front desk, at a glance, who is next, who is online or walk-in, who has replied, who has been flagged, and anything administrative such as an insurance form pending. Keep it to logistics. A clinical note should be written by the clinician in the record, not generated from patient texts.

A sample hand-off line, illustrative: "Walk-in, checked in 10:42, position 4, replied 'running to the car, back in 5 mins' (logistics), no flags." And a flagged line: "ONLINE, position 2, reply contains 'chest' and 'tight', FLAGGED, nurse alerted 11:03." The AI's job is formatting these consistently from the queue data. The rule it must follow: never describe or characterise a patient's condition in its own words, only quote what the patient wrote and mark the flag.

Writing the templates with AI: a prompt, a sample and the fixes

This is the one place AI does real writing in the project, so give it a tight brief. A clinician approves whatever comes out, and the prompt makes that review fast.

Write SMS templates for a walk-in clinic's patient-flow messages.
Rules: plain characters only (no emoji, no curly quotes), each under
160 characters, no medical advice, no mention of symptoms or
conditions, no promises about time. Use [first name], [clinic name],
[position], [range] and [number] as placeholders.
Write one version each for: check-in confirmation, wait update,
almost your turn, your turn, missed call. End each with a way to
opt out or reach reception, whichever fits.

A sample output (illustrative) for the check-in confirmation: "Hi [first name], you are checked in at [clinic name]. Place [position], estimated wait [range]. Times can change. Reply STOP to opt out." That is about 130 characters, so one segment. The usual fixes a manager makes are small but important. The draft may use "ETA", which some patients read as arrival time at the clinic rather than wait; change it to "estimated wait". It may say "we will see you shortly", a promise the queue cannot keep; delete it. It may add "we hope you are feeling OK", which invites a symptom reply; delete that too. Check the character count with the provider's tool after the placeholders are filled, since a long first name or clinic name can push a message into a second segment and double its cost.

A before-and-after shows the pattern. Before: "Hello, thanks for choosing us! Your wait is approx 20 mins, see you soon, hope it's nothing serious." After: "Hi [first name], you are checked in at [clinic name]. Place 3, estimated wait 20 to 30 minutes. Times can change. Reply STOP to opt out." The second version is longer by a few words and safer by a mile: it has no promise, no guess about the patient's condition and no curly apostrophe.

What a text should never reveal

Messages show on lock screens, on shared family phones and in notification previews. Keep the clinical reason for the visit, the service line and any result out of every text. "You are checked in at [clinic name]" is fine. "You are checked in for your sexual health screen" is not. The same rule covers names: use a first name only, or none at all, and avoid details that identify a child. If a patient texts a clinical detail, your reply must not repeat it back.

Ask your messaging provider how it stores message content and for how long, who at the provider can see it, and whether they will sign the privacy contract your compliance adviser requires for a vendor handling patient information. Our confidentiality checklist and the vendor security questions give a ready list. If any answer is vague, send less: a queue number and a name-free message need far less trust than anything else you handle.

Accessibility and the patients who will not text

Some patients have no mobile phone, share one with a relative, cannot read small text or prefer to be called. A queue system that only texts quietly pushes those patients down the line. Decide in advance how the front desk will handle them: a paper number, a verbal estimate at the desk, a screen in the waiting room that shows queue positions by number rather than name. Offer the same estimate by voice to anyone who asks.

Language matters too. If a large share of your patients read another language, have templates translated by a qualified person rather than a chatbot, and test them with native readers. A mistranslated "your turn" is harmless; a mistranslated "call the emergency number" is not. For patients who reply in a language you do not support, the catch-all rule from the previous section applies: the front desk sees it.

Texting costs at 90 patients a day

Suppose a clinic sees 90 patients a day and sends an average of four texts to each, so 360 messages. Twilio, for example, lists $0.0083 per outbound segment on its pricing page, plus a carrier fee of roughly $0.0035 to $0.005 a message, so one segment lands at a little over a cent. Other providers and destinations differ, so check the current list price instead of trusting a figure quoted in any tutorial. At an illustrative 1.3 cents a message, that is about $4.68 a day, or about $140 for a 30-day month before any platform fee. Two segments double it. This is why the 160-character discipline matters: trimming templates saves real money at volume.

What comes back is harder to price. If the front desk answers 25 fewer "how long is the wait?" calls and walk-up questions a day, at about two minutes each, that is roughly 50 minutes of staff time. If walk-outs before being seen fall by even two patients a day, the clinic keeps visits it would otherwise lose. Neither figure is guaranteed. Measure your own baseline for a week before launch so the comparison is fair.

Training the front desk to use the hand-off view

A tool does little if staff distrust it. Run a twenty-minute session before launch. Walk through the six templates, show the hand-off view with a flagged example, and have each person practise the red-flag response on a test message. Agree who owns the red-flag alert on each shift, with a named backup for breaks. Post a one-page card at the desk: what the system sends, what it never does, who to call if it misbehaves, and how to switch it off. The ability to pause every automatic message with one action matters on the day something goes wrong, such as a power cut or a data fault, so test that too.

Stage 7: roll out in three weeks and check it worked

A sensible rollout for a small clinic:

• Week 1. Write and approve the six templates and the red-flag list. Set up opt-in on the check-in form. Test internally with staff phones.
• Week 2. Run with real patients on one shift, with a manager shadowing. Record every message and every reply.
• Week 3. Review: estimate accuracy, complaints, stop requests, red-flag matches and false alarms. Adjust. Then widen to all shifts.

Judge it on four measures: the share of patients who received an estimate within ten minutes of check-in, the gap between estimated and actual wait, the number of walk-out patients who left before being seen, and the time from a red-flag match to a human responding. If walk-outs fall and the red-flag response is quick, it is working. If the estimate keeps missing by more than 30 minutes, fix the numbers before adding features.

Duplicate texts, after-hours replies and other edge cases

A few cases to plan for. A patient replies in another language: add a catch-all that sends any unrecognised reply to the front desk rather than ignoring it. A text arrives after closing: set an automatic reply with opening hours and the emergency number. A queue glitch sends the same message twice: add a rule to suppress duplicates within ten minutes. A patient texts "STOP" and then arrives: honour it, record it and tell them in person where they stand. Nothing here needs sophisticated software. It needs the rules written down before launch.

For reminder patterns on booked visits, which behave differently from walk-ins, see cutting missed appointments and, for the real monthly cost of texting, what SMS really costs. If you want help mapping both queues and writing the red-flag rule for your own clinic, that is what my AI implementation consultation covers.

Questions urgent care managers ask

Can the text tell a patient how long the wait will be?

It can show an estimate, framed as an estimate. Take the figure from your own queue system or a simple rule such as patients ahead times average minutes per patient, and say it may change. Never promise a time, because a sick patient arriving can push everyone back.

What if a patient replies with a symptom?

The system should not interpret it. Match a short list of red-flag words and phrases, and send any match to a named staff member at once, with an automatic reply telling the patient to call the emergency number if it is an emergency. Anything else goes into the queue for the front desk.

Do we need patient permission to text them?

Yes, get clear opt-in at check-in, record it, and include a way to stop messages. Business texting also has carrier and legal rules that vary by place, so ask your messaging provider and your compliance adviser what applies to you before sending.

Does this need a clinical safety review?

Messages about queues and logistics are lower risk than anything that interprets symptoms, but a clinician should still approve every template and the red-flag list. Keep a record of who approved what and when, and review the list after any near miss.

Further reads

• Best AI Scribe for Urgent Care: Fast-Visit Notes and Coding Checks — The clinical-notes side of urgent care, kept separate.
• How a Small Clinic Can Use AI at the Front Desk Safely — Front-desk AI rules for a small clinic.
• Is an AI Scribe Worth It for a Small Private Clinic? — Whether an AI tool pays for itself in a small clinic.
• 10 Admin Tasks a Small Clinic Can Hand to AI This Month — Ten admin tasks a small clinic can hand to AI.
• AI Answering Service for Home Care Agencies: After-Hours Call-Outs — Out-of-hours messages and call-outs, a related routing problem.
• AI Platform for Doctors: One Suite or Separate Scribe and Phone Tools? — One platform or several, for a clinic's AI tools.
• How to Train Employees on AI in a Clinic: Four Short Sessions — Four 30-minute sessions for clinic staff: rules, de-identified practice, consent wording and incident reporting, plus a sign-off sheet.
• How to Use ChatGPT for Data Analysis in a Clinic: No-Show Rates — Strip names, dates of birth and phone numbers, then ask ChatGPT for no-show rates by day, slot and lead time, and check the totals by hand.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Twilio's SMS character-limit glossary and SMS pricing page, Solv's provider pages and Experity's patient-engagement page (all checked 4 November 2026), and general guidance on patient messaging consent. Readers must confirm local messaging and health-privacy rules with their provider and adviser.