Skip to content
LLDTEK

Blog/Verticals

Four doors, four calendars. One shared inbox is where it breaks.

Same five jobs at every address. Hours, menus, and the stop person still have to stay local to the door that owns them.

Mateo Ruiz

Mateo Ruiz

Aug 29, 2026

Rain-soaked street of wooden shopfronts with lit lanterns and noren, each doorway its own shop

On a Thursday at 3:40, Market loses a colour. Mara is at Cedar, twelve minutes away, and she is not going to drive it. The chair is empty, the commission is empty, and every vendor she has sat with this month has offered her a group waitlist as if that were the adult way to run four doors. The people on Westfield's list booked a gel fill two miles from here. Cedar's list is treatment rooms. Lake's list is memberships on a twenty-one-day clock. Filling Market from any of those lists is how you turn one no-show into two unhappy visits and a review on the wrong Google listing.

The operating model for three to ten doors is the same jobs on every line, each job bound to that door's calendar, with one place for the owner to watch. Hours, menus, and staff lists look like duplication in a demo. They are how each floor makes money. Flatten them to make the software simpler and you will send a 7:40 text on Westfield's number to a sister door that is still lit, because someone decided the brand has one closing time.

The sequence for going live door by door is a different piece, Start with the door that already loses the phone, not the newest lease. Stay here for what has to stay separate once the jobs are on.

Keep hours, menus, and staff on the door that owns them

Westfield takes walk-ins until seven and runs Vietnamese and English at the desk. Nia manages that floor and she will tell you, without a dashboard, which holds she will never put through a script. Market is colour, dark on Monday, and will not put a toner in a forty-five-minute hole. Cedar has rooms with walls. Lake sells memberships, and Omar's next table is not a leftover from Cedar's 90-minute room. Mara owns all four. She does not sit any of those desks.

Those local facts are not settings you sync because the owner bought a group plan. They are the product at each address. A combo ticket at Westfield has a duration Market does not sell. Cedar cannot stack two bodies in one room because the grid looks empty during processing. Lake's member already has a date, and the rebooking job has to read that and stop. If the agent layer cannot keep those rules on the door that owns them, it is a shared inbox with a nicer logo.

What you want from the agents is narrower than a group calendar. Run the same five jobs at every door, with the same definition of done and different arguments the moment a number rings: front desk, recovery, rebooking, reputation, and floor. When Westfield's line hits, `who_is_open` reads Westfield, `book_visit` writes Westfield, recovery texts Westfield's waitlist, and floor shows Westfield's chairs. Market does not appear in that run.

A published number per door is how most groups this size keep language, hours, and the stop person honest. One mashed number is a choice you make on purpose, with a door check first, because otherwise Westfield's walk-in rules land on Cedar's rooms. If you already print four numbers, keep four numbers.

Local to that doorVisible to the owner
Hours, menu, roster, waitlistRuns, holes, and which door they belong to
VIP list and the named stop personHandoffs that still need a person
Google listing and published numberReview asks that fired or stopped
Language mix on that lineFilter by door, never one mashed calendar

Sharing a client record across doors is a POS question. Filling a 3:30 at the wrong address is an agent question. Those are not the same job, and treating them as one is how groups get a clean-looking workspace and a dirty Saturday.

What a shared waitlist actually does

A waitlist is not spare inventory for the brand. It is people who wanted a duration, at an address, often with a preferred person, in a language, with a commute they already made once. When recovery texts that list, it is asking one of those people to take a specific hole, once, in the language they already used, at the door that lost the visit.

Share the list and you change the question. Jules is on Market's list because she can take a 3:30 colour with Carly. Hanh is on Westfield's list because she wants a gel fill near the plaza, in Vietnamese, after work. Text both lists for Market's hole and one of three things happens. Jules says yes and you have a fill, which is the job. Hanh says yes, drives to Market, and sits in a colour chair she did not ask for. Or both say yes and you now have a double book plus a person who will tell their group chat the brand does not know its own shops.

The commission picture is worse than it looks in a utilization report. Carly lost a colour. Putting Hanh in that chair does not restore Carly's colour ticket. It creates a visit Market does not really sell, on a roster that is not staffed for it, and it teaches Westfield's waitlist that a text from the brand might mean a different door. The next time Westfield has a hole, those people will not trust the text. You burned two lists to look busy at one address.

The same break shows up in other trades with different nouns. OpenTable will fill a last-minute table from the waitlist at that dining room, and OpenTable for Groups can show sister availability when one room is full. That is a hospitality choice a GM can want. It is still not a reason for recovery at Market to text Westfield, and the hospitality front desk is at A hole at the wine bar is not a table at the noodle shop. A clinic cancel is a duration in a room that cannot overlap, which is why occupancy across two books is its own surface at Who is available across two clinic locations. A field cancel is only a fill if the next job can take the windshield time and the parts. Write sharing as an exception: a membership that may travel, a client who asked for any door, a four-top you would rather seat somewhere than lose. Do not let the owner view invent it because a merged waitlist looks tidy.

Bind the same jobs to different books

Five jobs travel. The sentences are short on purpose, because they have to be identical at every door. Front desk answers voice, SMS, and web, and books the person who is actually free on that book, including after hours. Recovery fills a hole from that door's waitlist. Rebooking puts the next visit with the same provider on the cadence that door already sells. Reputation asks once, in the client's language, against the Google listing for that address. Floor shows who is with a client, who is free, and which chair or room is dark, on that floor only.

What does not travel is the argument list. Westfield's front desk has to ask combo or single service because that is what Westfield sells. Market's front desk has to know colour cannot overlap a cut. Cedar's front desk has to refuse a 50-minute squeeze in a 90-minute room. Lake's rebooking has to read the membership clock. Copy those rules onto one mashed menu and you will land a 90-minute combo in a 45-minute chair because the agent found an open slot on a sister book.

Reputation is the quiet isolation failure. Google wants one profile per address. A good visit at Cedar should ask against Cedar's listing, once, in the language the guest used. Brand pages and flagship listings are how groups bury a door that is doing fine and inflate a door that is not. Door-level review work is at How to ask for a Google review after a salon visit. The agent follows the listing, not the brand.

Floor is the loud one. An open chair at Market is not an open room at Cedar. A tech who works Westfield on Tuesday and Market on Wednesday is a staff schedule, not a merge. If she is eighteen minutes into a visit at Westfield, Market must not see her as open. Square operators have already reported personal events leaking across location calendars. That is a POS bug. The agent should not repeat it.

You do not need five jobs on day one to run this model. You need the bind to be true for whichever job you staff. Front desk with a shared book is still a mashed queue. Recovery with a shared waitlist is still the 3:40 problem Mara already has.

Watch runs per location without mixing calendars

Mara does not need a fifth desk. She needs to see, without mixing four calendars, what already happened. Westfield booked three after-hours visits on its own next open hour. Market filled the 3:30 from Market's list. Cedar handed a VIP to Bea. Lake skipped a rebook because the member already had a date. She should be able to filter live runs, threads that still need a person, holes recovery is working, and review asks that fired or stopped, by door.

If the owner view starts to look like one inbox, two things go wrong at once. You cannot tell which number rang, so you cannot tell which stop person owns the thread. And Mara, who is good at this, will start taking desk work because the screen looks central. That is how an owner view becomes a call center. Nia still owns Westfield holds, Bea still owns Cedar rooms, Omar still owns Lake memberships. Mara watches. She does not become the overflow for four doors.

Status has to be honest on each run: live, done, or handed off. A transcript with no door, no tool call, and no named person is not a run you can staff. If recovery is already texting Jules for Market's 3:30, Nia should not also be calling Westfield's list from her pocket.

The till already owns the calendar

Keep the book you trust. Switching POS to get a smarter line is the wrong trade, and it is a longer argument at Do you need a new POS for an AI receptionist. Payments stay with the processor you already use. Four doors can even run four brands: Westfield on Vagaro, Market on Square, Cedar on Zenoti, Lake on Booksy. The agent still has to know which door rang.

What the till already finishes is calendar, checkout, payroll, cards on file, memberships, and rooms. Square can filter a calendar by location. Vagaro can share customers inside a group if you ask it to. Zenoti's Nearby Availability will suggest another center when one is full, which is a POS decision you turn on, not a default the agent should inherit. None of those products sit the line while every tech, host, or practitioner is in a service.

That gap is the agent layer. LLDTEK staffs the five jobs beside the book you already run. You do not prompt them, and you do not maintain a builder. The work is to listen to how each door actually runs, bind the tools to that calendar, and watch a real day with you. Staff front desk first, then recovery when the holes are the pain, then rebooking, reputation, and floor after the bind is true. Five agents on four doors on day one is how a group stops watching.

A website chatbot can collect a name and a preferred time. It cannot see who is free at Westfield, and it has no stop when a client asks to move a VIP hold from Lake to Cedar. An answering service can pick up and leave a note; Nia calls back when the floor lets her, and by then the walk-in is gone. Score those layers on last week's holes at What to test on an AI receptionist demo.

One door that already answers every ring, including after close, does not need this layer. Three to ten doors is the size where one owner can still name the stop person at each floor. Thirty doors is a call center.

Name the stop at the door that rang

A stop condition is how the model stays staffed. The agent finishes the job it was hired for, then it stops, and a person at that door takes the thread. Stop on a VIP hold, a dispute, anything off-menu including a hold moved from Lake to Cedar, an overlap, a request for a person the book cannot do, or a slot you marked as human. Nia names Westfield's VIPs. Bea names Cedar's. The list is per door. The agent does not negotiate, and it does not move a fight to a quieter address.

Most operators keep a person on the floor and let the agents take overflow at lunch, during a long appointment, on Saturday rush, and after close. That is coverage, not a firing plan. Mara sees the handoff: which door, which agent, which named person. If the workspace starts mixing calendars, isolation failed, and you pull the job until the bind is true.

After close is the test people skip. Westfield's line still rings at 7:40. Later hours at a sister door are irrelevant. After-hours voice books the person who is actually free on Westfield's book for Westfield's next open hour. It does not walk the caller two miles because the lights are on. VIP and off-menu still stop, even at 7:40, and they still stop to Nia, not to Mara.

Can one number serve every door

Only if you want it to. Most groups this size keep a number per door so language, hours, and the stop person stay local. If you already publish one number, staff it with a door check first. That is a choice you make while looking at the actual books.

Can a tech work two doors

Yes, if the POS already allows it and the hours do not overlap. Front desk books the door she is on. Floor shows her at that door only. If she is mid-visit at Westfield, Market must not see her as open.

What if one door already answers

Leave that door. Staff the ones that lose the walk-in and the after-hours call. Westfield can go live while Lake waits. Mara still sees the doors that are live.

Do we turn on every job at once

No. Bind front desk, watch it, then add recovery when the holes are the pain and the waitlists are still local. Five jobs on four doors on day one is how you stop seeing the run. The sequence is the companion post.

Updated Aug 29, 2026.

Next step

Thirty minutes against your book.

Bring last week’s no-shows and the POS you already run. If we are the wrong layer, we say so on the call. Numbers only if it is a fit.