What happens after your client signs an AI buildout

Three things happen between a signed deal and a live build: you collect intake under your brand, you order the delivery tier on the fulfillment platform and set your retail price, and the client moves into your portal while the build gets configured and tested. The part that decides how long it takes is not the build. It is how fast the client answers questions about their own business.

That is the honest version of the handoff, and it is worth knowing before you close your first deal, because the week after a signature is where most agency owners lose momentum. The prospect was excited on Tuesday. By the following Friday nobody has sent a logo and the enthusiasm has cooled. This post is what we do in that window, who owns which piece, and where it actually goes wrong.

What happens between signature and go-live

The sequence looks like this.

1. Move them off the demo. The live demo did its job. It is a sandbox, it is not their system, and nothing in it is connected to a real calendar. Once they sign, the demo stops being the deliverable and becomes the reference: this is the thing we are now building against for real.

2. Collect intake, under your logo. Onboarding is one branded path that gathers opening hours and holiday cover, the service list with a price band for each, brand assets, the calendar, the phone number, and who gets the leads. There is no account for the client to create and nothing to install. Whatever the form asks by default, you can change for your market.

3. Order the build. You pick a delivery tier on the fulfillment platform at wholesale and set what the client pays. There is no separate build invoice sitting in the middle of the deal. We covered the arithmetic on the retail side in how much to charge for an AI buildout, so I will not repeat it here.

4. Configure and test. The intake becomes what the agent knows. It sets what it can book, what it refuses to answer, and when it hands the call to a human.

5. Decide how calls reach it. More on this below, because it is the step people forget to plan and the one with a real external dependency.

6. Hand over the portal. The client gets a branded home with status, assets, and next steps. Your name is on it.

Steps 2 and 5 involve the client. Everything else does not. If you want a faster delivery, work on 2 and 5.

Why intake sets the clock

Look at what any voice agent needs before it can pick up a phone, and you find the same list everywhere. HighLevel's own documentation for building a voice agent asks for a defined use case, approved business information for the prompt or knowledge base, a calendar or workflow or transfer number for whatever action the agent takes, and an eligible phone number, and then walks you through testing transfers, actions, silence, and interruptions before you point real traffic at it (HighLevel support docs). ServiceTitan's setup for its roofing customers is the same shape: agent persona and knowledge base, opening and closing messages, custom follow-up questions, objection handling, which job types the agent may book, how it communicates the dispatch fee, and what triggers an escalation to a live person (ServiceTitan help center).

Read those lists again. None of it is engineering. All of it is business information that exists only inside the owner's head.

That is the real work of onboarding, and it is why "how long does delivery take" has no clean answer. Ask a four-truck HVAC company what they charge for a diagnostic visit and you will get three numbers from three people, one of which has been wrong since March. Ask a dental practice which appointment types a new patient may book without a callback and you will start an internal argument. The build is waiting on a decision the business has never had to write down.

Two things help. Get one named decision maker rather than a committee, and ask for artifacts they already have instead of asking them to compose new ones. Their current voicemail greeting, their price sheet, their intake script, the sticky note by the front desk phone. You are not writing a policy document. You are collecting what they already do.

Who owns what

You own the relationship, the retail price, and the account. You are the one who calls the client when something needs a decision, and your brand is on every surface they touch.

We own delivery. The fulfillment platform takes the order, the build gets configured against the intake, and onboarding assistance chases the missing inputs so you are not the bottleneck between signature and delivery.

The client owns the inputs and the go-live decision. Nobody can supply their hours, their prices, or their escalation rules for them, and nobody should be flipping their phone routing without them saying go.

The mistake I see most is an agency owner who thinks handing off fulfillment means handing off the client. It does not. A build that goes live without the owner having heard it once will get switched off within a week, no matter how good it is.

What about the phone number?

Two options, and the default answer is the boring one.

Forwarding is the safe one. Their existing number keeps ringing where it rings today, and unanswered or after-hours calls forward to the agent. Nothing about their carrier setup changes, nothing breaks if you turn it off, and you can start with a narrow slice such as evenings and weekends. ServiceTitan tells its own customers to start by forwarding a single tracking number before expanding, which is the same instinct.

Porting the main number is heavier and slower. FCC rules require simple ports, meaning generally no more than one line and no complex switching changes, to be processed in one business day, but the same guidance notes that porting from wireline to wireless can still take a few days (FCC porting guide). A small business's main line is often not a simple port, and a botched one takes down the phone that the entire company runs on. If your first delivery for a client is an outage on their main number, there is no recovering the relationship.

Start with forwarding on overflow and after hours. Prove it works on real calls. Talk about porting later, or never.

Where this gets slow, and who it is not for

Some plain admissions.

We do not publish a fixed delivery SLA, and I would be suspicious of anyone in this category who does. The configuration step is fast. The intake step depends on a business owner who is on a roof or in an operatory, and a promise of a guaranteed turnaround is a promise about somebody else's calendar.

Your first three deliveries will be slower than the rest. Not because the platform changes, but because you have not yet learned which four questions to ask on the close call so the intake does not stall. Once you sell the same trade repeatedly, you start collecting the answers during the sale instead of after it. That is also the argument for picking a lane early, which is what the roofing playbook is about.

Your buyer is probably buying AI for the first time. Census Bureau data through May 2026 put overall business AI use somewhere between 17 and 20 percent, with firms of fewer than 20 employees sitting under 20 percent and showing no significant increase over that stretch, while firms of 250 or more reached 37 percent (Census Bureau). The small service businesses you sell into are the least likely to have done this before. Onboarding a first-time buyer includes explaining what the thing will and will not do, and that takes time you should budget for rather than resent.

The monthly usage limits on your plan are for selling, not for a client's production traffic. The 100 voice minutes, 500 chat messages, and 50 audits a month on the $49 plan cover about forty demo calls, which is a sales allowance. Delivered client systems are ordered through fulfillment. Do not confuse the two lines when you are planning volume, and check the full limits before you promise anything.

If you want to own the codebase, host the infrastructure yourself, and configure every model call, this is the wrong product. You would be paying a monthly fee for a delivery path you intend to replace. Build it yourself and keep the control.

FAQ

How long does a build take after signature? The configuration is not the constraint. Assume the schedule is set by how quickly you get hours, services with prices, the calendar, and a routing decision back from the client. Agencies who collect most of that during the sales conversation move much faster than agencies who start asking after the signature.

Do I have to be on the fulfillment calls? You stay the point of contact with your client. Onboarding assistance exists so the collection of inputs does not fall on you and so a missing logo does not become a three-week delay, but the client relationship stays yours and the client never deals with us directly.

What does the client see during all this? A branded portal with status, assets, and next steps, under your agency's name and colors. Prospects and clients do not see Lumina on a published demo. The one thing always visible on a demo page is the sandbox disclosure stating that the AI is a demo and that bookings in it are simulated.

What if I stop paying? Nothing is deleted. Published demos unpublish so the links stop serving, everything you built stays as a draft, and you stay signed in with zero voice, chat, and audits until you resubscribe. Then it goes back up unchanged.

Do this before your next close

Write down the four questions your trade always fumbles. For HVAC it is usually the diagnostic fee, the emergency versus maintenance split, who gets the lead text, and which calendar is real. Put them in the close call, not the follow-up email, and you will have removed most of the delay before the contract is even signed.

Then go talk to the demo and listen to what it asks for. That list is your intake form.

If you have been waiting to start, the plan is $49 a month and the posted price rises on October 1.