AI Lead Qualification That Decides a Job Before a Truck Rolls

The Lead Is Not the Unit of Cost. The Slot Is.

A weak lead does not cost a service business an hour of selling time. It costs a technician's day, a slot on the board, and the drive to an address that was never going to become repeat work. The scarce resource is not leads. It is capacity, and every unqualified job that gets booked spends capacity that cannot be resold.

This is why "you never miss a lead" is the wrong virtue for a business that runs on trucks. Universal coverage is a marketing promise from a vendor whose cost of a missed lead is zero. Your cost of a bad one is a slot. The two businesses are optimizing different things.

Every intake decision runs against three constraints you already know and already write down somewhere: the trades you hold, the area you cover and the drive it takes to reach it, and the minimum job worth dispatching. None of those require a prediction. They require a rule and a record to check it against.

A filter

Runs a rule against a record and produces yes, stage, or no.

A score

Ranks leads after the fact and changes no truck roll.

The decision

Made at intake, against capacity, before a slot is spent.

The Intake Fit Model

Move the sliders to match your intake. The model separates leads into servable, staged for a person, and declined, then shows what the declined share would have cost in truck rolls. The honesty dial is the below-minimum share, because a caller rarely knows what a job costs the business to run.

Keep the human-decision dial above zero until your own numbers say otherwise. A record cannot settle every call, and the model is built to say so.

Where the leads land

Capacity recovered, not cash

96 rolls / month

about $213,120 a year in truck-roll cost, and this is capacity rather than cash. It is worth the figure only if you refill the slots with work you can serve.

Illustrative estimate from the inputs above. Your own numbers will differ; this is not a measured result.

Lead types by what a record can settle

The property address against the published service area on the platform

The cleanest screen there is. An address is a field, and a map is a rule. No judgement is involved.

The asset or vehicle field against the trades you actually hold

The platform already knows what is at the property if anyone ever entered it. Most have not.

Job type, estimate history, and typical ticket at that property

The same words describe a fifteen minute call and a three hour diagnosis. The record tells them apart.

The contact and decision-maker fields, unevenly stored

The caller says the owner will decide. Whether that is on the record varies by platform and by account.

Nothing. No platform stores real urgency

This is the signal a scoring model leans on hardest and the one no record can confirm. It routes to a person.

Weighted, about 70 percent of this mix is something a record can settle without a person. Compare that against your own human-decision dial.

One Lead, Six Stages

Each stage names the system it touches, the payload it leaves behind, and the concrete failure it prevents. The flags are the human boundary made visible.

Target System

GoHighLevel contacts and conversations

Lead Arrives and Is Attributed

The lead lands with a source and a timestamp. Before anything is decided, the system knows where this came from, because a decline is only useful if it can be traced back to the source that produced it.

The lead arrives with no source, so the eventual decline teaches nobody anything about the channel.

Stage payload

What Callers Get Wrong on Purpose

Four intake signals a caller has a reason to shade, and why each one is better read from a record than from the call.

The Real Need

Callers describe symptoms more readily than causes, and sometimes they describe a cause they were told by someone else. The affordance question (what does the record show this unit is) beats the motive question (what do you think is wrong).

The Urgency

Everything is urgent until a window is offered. Urgency is the single most weighted input in a scoring model and the one no operational platform stores. It is worth a person, not a number.

The Budget

Asked directly, a caller will name a number to end the question. Asked what they had done to the property before, the record already shows what they actually spend. History is a better budget question than budget is.

The Authority

The person on the phone is often not the person who says yes. If the record does not show decision authority, the call needs a human, because a quote delivered to the wrong person is a quote that expires in silence.

"I am just getting prices" is not a disqualifier. It is a decision signal. It says the caller needs a number before they commit, which is almost every caller, and it says the visitor wants that number to be honest. A business that cannot see decision authority on the record needs a person on that call, because the alternative is a quote sent to someone who will read it and do nothing.

Two Things We Will Tell You to Buy Elsewhere

Not a lead-scoring model

A scoring column ranks leads against each other and produces a number nobody trusts. It changes no truck roll, because a score is computed after the decision was already made. A fit decision is made at intake against rules you can write down and a record you can check. If a tool cannot tell you why it produced its number, it is a score, and your dispatchers will ignore it.

Not a robot that turns callers away

A decline is a routed outcome with a reason and an alternative, and a person sees it before it goes out. A system that turns callers away on its own is not qualification. It is a rule that loses customers you never got to judge, which is worse than a lead a person read and declined on purpose.

Where the Line Sits on Every Intake Decision

Qualification sits next to money, so the boundary matters here more than almost anywhere. The system gathers the evidence and proposes. A person decides who gets turned down.

Autonomous

• Read the lead, its source, and its timestamp
• Match the caller to a person, a property, and any asset on file
• Check the address against the service area
• Check the trade against the work you perform
• Compare the stated need against the posted minimum
• Propose a verdict of servable, staged, or declined with a reason

Supervised

• Draft the response that offers an alternative to a declined lead
• Stage a proposed appointment window without committing to it
• Flag a claim that contradicts the record for a person to confirm
• Route an authority-unknown caller to a human rather than a quote

Never

• Quote a price before the work has been scoped
• Confirm an appointment or promise an arrival time
• Promise coverage outside the published service area
• Make a warranty or scope exception on its own
• Decline a caller without a person seeing the reason
• Record a caller's claim as a verified fact

Action, Tier, and the Reason Behind Each

Straight Answers

Where This Fits in the Lead Lifecycle

This page owns the decision about whether a job should exist. The pages around it own everything else.

Send Us Your Last Thirty Leads

We will mark each one as servable, staged, or declined, with the reason written next to it, so you can see how much board capacity went to jobs that were never a fit. You keep the analysis whether or not you hire us.