Log two weeks of service start and end times and have an AI assistant turn them into real durations per service. Then match each walk-in to the qualified technician who frees up first, estimate the wait from the work in progress and the queue ahead, text it as a range, and let the desk override it.
None of this needs a nail-design generator or a special booking app. It needs three tables you write down once: how long each service really takes, who can do what, and how the queue is served. An AI assistant helps you build the tables from your own data and draft the texts. It should not be left to guess durations from general knowledge, because a salon doing quick gel manicures and one doing long acrylic sets have different waits.
What a wait estimate actually is
A wait estimate answers one question: when will a technician who can do this service be free to start on this person? Four things set the answer. The work already under way (each technician has a service that ends at some time). The queue ahead (clients who arrived earlier and have first call). Who is qualified for each service. And the reset time between clients, which in a nail salon means clearing the table, cleaning tools and wiping down, so it is rarely zero.
Four misconceptions cause most of the early disappointment:
- "The AI predicts the wait." It calculates one, from the durations and rules you give it. With no durations of your own, any figure it produces is a general guess.
- "A wait estimate is a promise." It is a range with a stated edge. The salon can beat it, but should not quote a single number as if it were fixed.
- "Matching means picking the best technician." Matching here is more basic: filter to the technicians who can do the service, then take the one who is free first. Preference and turn rotation are policy, set by you.
- "A busy salon needs a complex system." The board on the wall, a sheet, and one rule written in plain words do most of the work. The assistant's job is to write and test the rule.
Barbershops run a simpler version of this queue, covered in the tutorial on walk-in queues, waitlists and texts for barbers. A nail salon differs because services run from 30 minutes to nearly two hours and because not every technician can do every service. That is what makes duration data and a skill table necessary.
Collecting service durations: a two-week timing log
Most booking and till systems record when a client was paid, not when the service started. If yours records start and end times, export two weeks of tickets. If not, keep a simple log for two busy weeks: a sheet or a clipboard where each technician writes service, add-ons, start time and end time. It adds about ten seconds per client.
Keep names and phone numbers out of the log, and out of anything you paste into a chat assistant. Service, technician level and two times are all the assistant needs.
I run a nail salon. Attached is a two-week log: date, service, add-ons, technician level
(senior or trainee), start time, end time.
1. Calculate duration in minutes for each row. Flag any row longer than 180 minutes or
shorter than 10 as a probable logging error and exclude it. Tell me how many you excluded.
2. For each service and technician level, give: number of records, median duration, and the
range that 8 out of 10 records fall inside.
3. Flag any service with fewer than 10 records.
4. Output a table I can paste into a spreadsheet. Do not explain causes.An illustrative result from 312 services (some bands built on thin data):
| Service | Records | Median (min) | Usual range (8 in 10) | Note |
|---|---|---|---|---|
| Basic manicure | 41 | 35 | 30-40 | Trainee median 40 |
| Gel manicure | 96 | 45 | 40-55 | |
| Gel removal and new gel | 52 | 60 | 55-70 | |
| Pedicure | 58 | 50 | 45-60 | |
| Acrylic infill | 38 | 60 | 50-75 | |
| Acrylic full set | 22 | 90 | 80-110 | |
| Nail art add-on | 5 | n/a | n/a | Under 10 records: use your own estimate |
Three fixes are worth making before you rely on the table. First, make sure the assistant has split senior and trainee technicians, because a trainee's basic manicure in this example runs about five minutes longer, and a pooled median hides that. Second, check how it handled combined services such as gel plus art. If the art sits in an add-on column, ask for the add-on time separately rather than letting it vanish into the average. Third, recompute one row by hand. Pick the pedicure rows in your sheet, take the median yourself, and make sure you land within a minute or two of the assistant's figure.
Your menu probably already lists durations that are guesses. The tutorial on writing a salon service menu and price list with AI shows how to put the corrected figures on the menu once the log has replaced the guesses.
The technician skill table and four matching rules
Write who can do what as a grid, with names replaced by letters if you share the file. Be strict: a service belongs to a technician only if you would happily book them on it tomorrow.
| Service | Tech A | Tech B | Tech C | Tech D (trainee) |
|---|---|---|---|---|
| Basic manicure | Yes | Yes | Yes | Yes |
| Gel manicure | Yes | Yes | Yes | No |
| Gel removal and new gel | Yes | Yes | Yes | No |
| Pedicure | Yes | Yes | Yes | No |
| Acrylic infill | Yes | Yes | No | No |
| Acrylic full set | Yes | Yes | No | No |
| Nail art | Yes | No | No | No |
Then write the matching rules. These four cover most situations, and each one stops a specific mistake:
- Skill gate. Only technicians marked "Yes" for the service can be assigned. This stops a trainee from being booked on a full set because they happen to be free.
- First come, first served. Walk-ins are served in arrival order, unless the skill gate forces a swap (a basic manicure may go ahead of an acrylic infill if only the acrylic technician is busy). Write this down so nobody quietly reorders the board.
- Earliest free. Each queued client goes to the qualified technician who will be free soonest, counting a five-minute reset after every client.
- Fit-in. Never assign a walk-in to a technician whose booked appointment starts before the walk-in's high-end finish, plus the reset. A booked client who arrives on time should not be kept waiting by a walk-in who ran long.
A fifth rule is for regulars. If a client asks for a particular technician, the message gives both options: wait for them, or start sooner with someone else. The client decides, and the desk logs it. Some salons also run a turn rotation to share clients fairly among technicians. If yours does, that rotation sits above rule 3 and the assistant must respect it. Do not ask the assistant to design pay-affecting rotation; that is a decision for you and your team.
Working the estimate by hand: a 2pm board, step by step
This is the calculation the assistant, or a sheet, performs. The queue board behind it needs only six fields per client: a ticket number or first initial (never a full name if the board goes into a chat assistant), service with any add-ons, arrival time, requested technician or "any", status (waiting, in service or done) and start time. The estimate reads nothing else, which keeps it easy to check and keeps client details to a minimum.
Suppose it is 2:00pm, the salon has four technicians (skills as above), and the board looks like this. Durations are the medians from the log, with the trainee's basic manicure at 40, and a five-minute reset after each client.
| Item | Service and duration | Technician | Starts | Ends | Free again |
|---|---|---|---|---|---|
| In service | Acrylic full set, 90 (started 1:10) | A | 1:10 | 2:40 | 2:45 |
| In service | Gel removal and new gel, 60 (started 1:35) | B | 1:35 | 2:35 | 2:40 |
| In service | Pedicure, 50 (started 1:30) | C | 1:30 | 2:20 | 2:25 |
| In service | Basic manicure, 40 (started 1:50) | D | 1:50 | 2:30 | 2:35 |
| Queue 1 | Gel manicure, 45 | C (earliest free) | 2:25 | 3:10 | 3:15 |
| Queue 2 | Pedicure, 50 | B (earliest free) | 2:40 | 3:30 | 3:35 |
| Queue 3 | Basic manicure, 40 | D (earliest free) | 2:35 | 3:15 | 3:20 |
| New walk-in | Acrylic infill, 60 | A (only A and B qualify; A frees first) | 2:45 | 3:45 | 3:50 |
The new walk-in's wait is 2:45 minus 2:00, so 45 minutes. To turn that into a range, run the same board twice more with every service at the low end of its usual range and then at the high end. Low: everything finishes early and the wait is 35 minutes. High: everything runs long and the wait is 65 minutes. The central estimate is 45, and the high case is 65.
What the client sees is not 35 to 65, which is too wide to be helpful. The rule is to quote from the central estimate up to about 15 minutes above it: "45 to 60 minutes." People are pleased when a wait ends early and annoyed when it runs over, so the range leans up, and any drift beyond the top gets a text update. The assistant should print the steps as a table like this one, so that the desk lead can check the single line that matters, the earliest free qualified technician, without redoing every sum.
A different walk-in shows the fit-in rule at work. Suppose the new arrival wants an acrylic full set (90 minutes, up to 110) instead of the infill. Technician A would be free at 2:45, but A has a booked appointment at 4:00 and a high-end finish of 4:35 clashes with it. Technician B is free at 3:35 after the pedicure in the queue and has nothing booked until after closing, so the estimate is 3:35 minus 2:00, or 95 minutes, quoted as "95 to 110 minutes". The right message there adds an alternative: "A booked time later today or tomorrow may suit better. Would you like options?" Some clients will take it, and the queue gets shorter.
The text-back: four messages and the wording rules
The text is where the estimate meets the client. Four messages cover the visit. Use first names only, plain characters and one clear action in each:
1. Confirm: "Hi [first name], wait for a gel manicure at [Salon] is about 45 to 60 min. Reply YES to hold your place, NO to cancel. We will text 10 min before your turn."
2. Update: "Update from [Salon]: your wait is now about [X] to [Y] min. Reply NO if you would rather leave."
3. Heads-up: "You are next in about 10 min, [first name]. Please head to [Salon] now. We will hold your place until [time]."
4. Missed: "We could not reach you at [time]. We will hold your place for 10 more minutes."Check the length when the names are merged. The first message is about 155 characters with the brackets in place and can tip past 160 with a long first name and salon name. Many text services bill messages in parts of 160 characters, and an emoji or a curly apostrophe can cut a part to 70, so test with your longest names, keep to straight punctuation and leave out the emoji.
Only text people who gave you their number for this visit, and for this purpose. A number collected to hold someone's place in the queue is not agreement to receive promotions, and the rules on texting differ by place, so ask your adviser before you reuse a queue list for marketing. If your texts go through a booking app, check how it counts them. Some apps and AI receptionist add-ons count texts as credits, and a message that runs past 160 characters can use more than one, so a busy salon texting each walk-in two to four times can use credits faster than it expects: 30 walk-ins a day at three texts each is 90 texts a day, or about 2,250 in a 25-day month, before any message runs long. The vendor's current pricing and help pages are the place to confirm the rules for your plan.
Manual override at the desk, with reason codes
The desk lead can see what no sheet can: the client at the dryer who is minutes from finished, the technician who has just gone for a break, the group of four who have walked in together. The rule is that the override always wins, and always leaves a note. A short code is enough:
| Code | When the desk uses it | What changes in the estimate |
|---|---|---|
| EARLY | A technician will finish sooner than the model thinks | Subtract the minutes you can see |
| LATE | A service is running long (a repair, art, a late decision) | Add the minutes you can see |
| BREAK | A technician is on a break or leaving early | Mark them unavailable until a stated time |
| GROUP | Several people arrive together and want to sit side by side | Estimate for the group: all the qualified technicians at once |
| REG | A regular asks for a named technician | Offer both: wait for them, or start sooner with someone else |
On the 2pm board, suppose the desk lead notices that Technician A's client is already at the dryer and A will finish at 2:30, not 2:40. The code is EARLY with 10 minutes, the wait for the new acrylic infill drops from 45 to 35, and the text to the client says "35 to 50 minutes". The note costs five seconds and earns its place later: after two weeks, paste the override log (codes only, no names) into the assistant and ask which codes recur and at what times. If EARLY appears on every Saturday afternoon, the durations for that period need correcting. If LATE clusters on art add-ons, the add-on time is too low.
The override also covers fairness. If an assistant suggests reordering the queue because a high-spend service would raise takings, the answer is no. First come, first served is a promise to people standing at the desk, and the only automatic exceptions are the skill gate and a booked appointment. The assistant is told that in writing.
Where the sum runs: a sheet, a chat assistant or a booking app
There are three places the calculation can live, and they suit different salons.
| Option | Cost | Strength | Risk |
|---|---|---|---|
| A spreadsheet with formulas the assistant helped you write | Free or the cost of the sheet | The same answer every time; the desk can see each step | Takes a few evenings to build and a person must maintain it |
| A chat assistant fed the board each time | Free tier or a paid plan at about $20 a month | Flexible; good while the rules are still changing | Can slip on arithmetic; always ask for the table of steps and check it |
| A booking app's own AI or queue feature | Add-on and message credits; check the current pages | Already connected to the diary and texts | Mostly sold for answering and booking, so test it against your own board |
My suggestion is to start with the chat assistant for two weeks, because the rules will change as soon as the first real queue tests them, and then turn the stable rules into spreadsheet formulas. Ask the assistant to explain each column as it writes the sheet, so that you can maintain it without calling anyone. Whichever you choose, the rule about live use is the same: if an assistant is doing the sum, make it print the steps in a table, and have the desk lead check the one line that matters.
The wider job list for the desk is in the tutorial on what AI can take over from a nail salon receptionist, and the cleaning log that feeds your reset time is covered in the nail salon cleaning checklist with daily logs. If late arrivals and no-shows are eating the gaps you hoped walk-ins would fill, cutting salon no-shows with reminders and deposits is the next place to look.
Three weeks from paper log to live estimates
Do it in this order, and resist the temptation to switch on texting first:
- Week 1. Keep the timing log. Write the skill table. Do not change how you run the desk.
- Week 2. Build the duration table with the assistant and make the three fixes. Write the four rules. Run the estimate by hand on the board at 11am, 2pm and 5pm each day and write the estimates down, then compare them with the real start times.
- Week 3. Start sending the confirm text and log every override. At the end of the week, work out how many walk-ins started within 10 minutes of the central estimate.
A target of seven in ten walk-ins starting within 10 minutes of the central estimate is a fair first goal. If you are below that, look at the codes before you blame the model. A cluster of LATE codes means the durations are too short, a cluster of BREAK codes means breaks need to be on the board, and a cluster of GROUP codes means the group rule is missing. If you are above nine in ten, the estimate is doing well and you can narrow the 15-minute margin.
Keep the whole thing humble. The estimate is a courtesy to the person who walked in, not a contract, and the desk lead's judgement is the final check. The assistant's job is to make that judgement faster, to turn two weeks of times into a table, and to stop the board from depending on whoever is on the desk that day.
Questions nail salon owners ask about wait estimates
How much timing data do I need before the estimate is useful?
Two busy weeks of start and end times for each service usually gives enough records for the common services, a few hundred in a salon with four technicians. Any service with fewer than ten records should keep your own estimate until more data builds up. Log the senior and trainee technicians separately, because their times can differ.
Should the wait estimate be one number or a range?
A range, weighted upwards. Give the central estimate and add about 15 minutes, such as 45 to 60 minutes. Clients are pleased when a wait ends early and annoyed when it runs over, and a range gives the desk room when a service overruns. If the real wait drifts past the top of the range, text an update rather than leaving people guessing.
Can the AI decide which technician gets which client?
It can apply rules you write: who is qualified, who frees up first, whether the job fits before a booked appointment. It should not decide turn rotation, pay effects or favouritism, which are policy for the owner. The desk lead can always override, as long as the override is logged with a reason code.
Do I need a paid AI receptionist add-on for this?
No. A shared sheet, a chat assistant to help you build and test the rules, and a text-sending tool you already use are enough to start. Add-ons are mostly sold for answering messages and booking appointments, so ask for a live demo of a walk-in wait calculated from service times and technician skills before you pay, and check how message credits are charged.
Further reads
- Best AI Receptionists for Salons, Compared by Price — Salon AI receptionists compared by price.
- Fresha AI Concierge: Is It Worth It for a Small Salon? — Whether one booking app's AI concierge pays for itself.
- Vagaro Pricing for a Small Salon: Calendars, Add-Ons and AI Features — Vagaro calendars, add-ons and AI features priced out.
- How Salons Use AI to Rebook Clients and Fill Gaps in the Diary — Rebooking and filling gaps in the diary.
- Barbershop Booking App: Walk-In Queues and AI Replies Compared — How booking apps handle walk-in queues.
- AI Voice Agent for Salons: Booking Rules to Set Before It Answers — Booking rules to set before a voice agent answers.
- Can an AI Receptionist Book Gel, Acrylic and Fills at a Nail Salon? — Whether an AI receptionist books gel, acrylic and fills correctly.
- Restaurant Waitlist Software: Where AI Helps at the Door — AI can quote waits by party size, text when a table is ready and answer waitlist calls. Here is what it still needs a host at the door to do.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Twilio's documentation on SMS character limits and encoding (check your own provider and booking app for how texts are counted). All service times, queue numbers and prices in the examples are illustrative.