Live in five steps. Days, not a rollout.
LLDTEK goes live in five steps: connect the calendar or POS you already run, configure what the agents may and may not do, test real scenarios against your book, launch beside your team, then improve from live conversations. You bring the operating knowledge. We do the build, so you never write prompts.
What goes live
Five agents, one thread.
They share one conversation across voice, SMS and the web, so a client never repeats themselves when the channel changes. A go-live rarely switches all five on at once. We start with one job on one channel, prove it against your book, then add the next.
- Front deskAnswers when nobody can leave
- RecoveryFills the empty slot
- ReviewsAsks for the review once
- RebookingBooks the next visit
- Live floorWho is open, right now
The five go-live steps

1
STEP 01
Connect
Link the calendar, POS, and records the agents need to see, and set the actions they are allowed to take.
- What you provide
- Access to the booking system or POS you already run, such as Vagaro, Square Appointments, Booksy, Zenoti or Mindbody. Your service menu with durations, the providers who perform each service, and any room, chair or table a service occupies. Opening hours for each location, including what changes after close.
- What LLDTEK does
- We map services, providers, resources, locations and durations into the agents’ approved tools. The read path is tested first: slots that should exist, slots that should not, and requests that cross closing time. Writes are switched on only after the reads come back correct, and each agent gets only the permissions its job needs. Checking availability, for example, does not require the right to cancel.
- What done looks like
- The agents see your real availability, a correct refusal comes back as reliably as a correct offer, and your calendar stays the source of truth. Nothing is migrated, and checkout, payroll and the till stay where they are.

2
STEP 02
Configure
Set the goals, rules, knowledge, actions, handoff paths and stop conditions for each agent.
- What you provide
- A representative week of calls, texts and booking questions, including the ones that were missed. The answers your best desk person gives without thinking: prices, policies, deposits, and what a new client needs before booking. A named person or role who owns each kind of exception, and how quickly they can realistically reply.
- What LLDTEK does
- We listen to that week and write a one-page operating brief for each job, covering the trigger, the goal, the inputs it needs, the actions it may take and the conditions that stop it. Phrases customers actually use, like “the usual” or “a table outside,” are mapped to bookable services, with one clarifying question where the mapping is unsafe. Handoff rules are written as observable conditions, such as a disputed charge or a service that is not on the menu, and each one is tied to an owner.
- What done looks like
- Your team can read the brief and agrees with every rule in it. If they cannot agree, the agent is not ready, because automation only makes a disagreement travel faster. Every stop has an owner and a sentence the agent says when it hands over.

3
STEP 03
Test
Run real scenarios against your book before the agents touch live demand.
- What you provide
- The edge cases your team does not want an agent to guess: the VIP who rings the main line, the off-menu request, the service that runs past close. The person who knows the calendar and the person who takes exceptions should both look at the results.
- What LLDTEK does
- We run one ordinary booking, then the cases that make a desk pause. The preferred provider is free but the room is not. Two requests chase the last slot. A complaint arrives on the same number as routine bookings. After every write we open the destination system and compare service, provider, location, start time and duration with what was confirmed. Confirmations are built from the record the calendar returns, not from what the agent meant to write.
- What done looks like
- Each case ends in a booking, a refusal or a handoff that follows the written rule, and each outcome appears in the run history. The ordinary and edge cases are saved so they can be run again after any later change to services, rules or integrations.

4
STEP 04
Launch
Go live beside the team and systems you already run, starting with a narrow window.
- What you provide
- A launch window with a real gap and a manageable risk. After-hours traffic or lunch overflow is often easier to isolate than a full Saturday. A named person watching handed-off threads during that window, and an agreed pause procedure: who can stop new calls, messages or calendar writes, and what customers hear while it is paused.
- What LLDTEK does
- We switch on the agreed job on the agreed channel and leave your phone line, POS and staff as they are. The five agents share one thread across voice, SMS and the web, so a call that continues as a text keeps its context. Before anyone is billed, we watch a real day on your floor with you.
- What done looks like
- Bookings land correctly in your calendar. Handoffs reach their owner with what the client asked, what was checked and what should happen next. The person at the desk still runs the floor, and the agents cover the hours they cannot: lunch, a full floor, after close.

5
STEP 05
Optimize
Review live conversations and outcomes: what booked, what dropped, and what still needs a person.
- What you provide
- A short review at the same time each day during the first week. Fifteen focused minutes is more useful than a weekly meeting built from memory. Corrections as you spot them, such as a stale schedule, a changed price or a new policy.
- What LLDTEK does
- We sort runs into completed, refused, handed off, corrected and abandoned, then trace each failure to its nearest cause: source data, service mapping, policy, the tool itself, handoff ownership or scope. We fix that source instead of piling more instructions onto the conversation. Once the current scope stays clean, we add the next channel, service, location or agent one at a time and test each as a new surface.
- What done looks like
- The correction queue shrinks, your team can explain every repeated handoff, and routine runs stop needing attention. You keep a one-page record of the job, channels, hours, permissions, stop conditions, handoff owners and pause procedure.
Every go-live starts with a thirty-minute call against your own book rather than a fictional business. Bring the calendar or POS, two representative services, one recent missed call and one case your team would never want an agent to guess. We run the ordinary job, run the case that must stop, and open the destination system to check what actually changed.
The call ends with a fit decision. If the agents match your hours, languages and book, we agree the first job, channel and owner. If you already answer every ring, are shopping for a new POS, or want to build and prompt agents yourself, we say so on the call.
See the systems we connect to
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.
