Say you run a solar sales org with 12 reps. Call it Meridian Solar. At 8:47pm on a Tuesday, a homeowner fills out a Facebook lead form with a roof photo, their utility company, and their average bill. Three other solar companies bought the same ad slot in the same zip code. Whoever calls first usually gets the appointment. By the second call, the homeowner has already told someone else their roof type and bill range, and most people won't sit through that conversation twice.
Meridian's leads wait in a queue until Wednesday morning, when a setter works through Tuesday's backlog in the order it arrived. Meridian has a routing problem, and fixing it doesn't take another hire.
What speed to lead measures
The number to track is the gap between when a lead is created and when someone first tries to reach it, using your CRM's own timestamps. Count the attempt whether or not anyone picks up. A missed call at 90 seconds beats a connected call at nine minutes, because the missed call puts your name on the homeowner's phone before the next company's does.
Set a firm target. The first outbound attempt goes out within 90 seconds of the lead landing in the CRM, every hour of every day the ads are running. If the ads run until 9pm, the response system runs until 9pm too. A form submitted at 8:47 that gets a callback at 8am the next day is close to twelve hours late.
The trigger: send leads to the CRM instead of someone's inbox
At most solar orgs, the delay comes from the path a lead travels before any setter sees it. Facebook lead ads, Google Lead Form ads, and the website's contact form each send someone an email notification, and that email waits until the person opens it. Speed leaks out at every step of that chain.
Replace the chain with a webhook. The ad platform posts the lead into the CRM the moment it's submitted, and the CRM fires an automation in the same second. Nobody has to notice the lead or forward it. An AI agent takes the first pass from there. Within the 90-second window it texts the lead. The text gives the agent's name and company and asks one question to start qualifying. It texts instead of calling because most people screen an unknown number at 8:47pm, but they'll read a text.
Hi {{first_name}}, this is Ava with Meridian Solar. Got your info from
the form you just filled out about solar for your home on {{street}}.
Quick one: is the roof owned by you, and is it a shingle roof or
something else (tile, metal, flat)?
That one message confirms there's a real person behind the lead and starts qualifying before anyone on your team touches the file.
The qualifying script: three facts, and drop bad fits early
An AI agent qualifying a solar lead doesn't have to sound like a rep reading from a script. It needs three facts it can check, gathered in whatever order the conversation takes.
Roof ownership and type come first. Renters belong in a different pipeline from a straightforward reroof-and-solar candidate, and so do roofs too old or too shaded for panels. Some of those leads belong in no pipeline at all.
Next is the utility provider and a rough monthly bill. A $60/month bill in a low-rate utility territory rarely pencils out. A $280/month bill almost always does. The agent can route the two differently before a rep spends 30 minutes on a call that was never going anywhere.
The last fact is the timeline, along with who's making the decision. The person texting back might be the homeowner, or a spouse who needs to bring someone else in before anything gets booked.
Bad-fit leads shouldn't vanish without a word, and they shouldn't take a slot on a rep's calendar either. Send them to a lower-touch follow-up sequence, an email plus a monthly check-in text, and keep the qualified leads moving toward booking.
The booking handoff: put it on a calendar
Most speed-to-lead setups break at this last step, and often nobody notices. The AI qualifies the lead well, then tells the homeowner "a rep will call you," and the wait the agent had removed comes right back. The agent should instead offer open slots from reps who cover that lead's territory and book the appointment itself. Then it texts a confirmation with the rep's name and photo.
The rep starts that call already knowing the roof type, the bill range, and the timeline. The qualifying texts serve as prep notes, so nobody has to read a transcript while the homeowner waits on the line.
This is also where you decide how much the AI owns and how much a person does. Some orgs want a human setter to review every booked slot before it's final. Others let the agent book directly and deal with the occasional overbooking as it comes up, rather than slowing the whole system down to prevent it. Both approaches work. You're trading speed against a small amount of rework, so make that call on purpose instead of letting the default make it for you.
What it costs to run
You don't need a call center for this. You need a CRM that can receive a webhook and an automation that sends a text within seconds of the webhook landing. Add an AI agent connected to both that can hold a short qualifying conversation and read a real calendar, and you're covered.
Building it takes something closer to a focused week than a quarter. The case studies on our site show real build times for comparable AI agent work. Solar isn't the same as those industries, but the mechanics carry over whatever you sell. A webhook brings the lead in, the agent qualifies it, and a calendar slot gets booked.
After launch, the ongoing cost is mostly someone's attention. Once a week, a person on your team should read a sample of the AI's conversations and fix the script wherever it answered badly. Skip that and the agent drifts. The 8:47pm text will keep going out on time, while the qualifying questions slowly stop matching what your reps need to know before an appointment.
Getting named when someone asks an AI assistant who to call is a separate problem from what happens after they fill out your form. The first post in this series covers that AI search visibility question. Speed to lead won't bring in new demand. It pays off by keeping the leads you already paid for from going to someone else.
FAQ
Does 90 seconds matter, or is 5 minutes close enough?
Treat 90 seconds as the target you design around. Nothing breaks at 91. The goal is to be first, and when three competitors bought the same lead, first place is usually decided within single-digit minutes. Build for 90 seconds and even a slow day lands you well inside five.
What happens when the AI gets a question it can't answer?
It should hand the conversation to a person right away instead of guessing. Questions about financing terms or HOA rules deserve a correct answer, and so do questions about a specific incentive program. The agent is there to triage and book, and it shouldn't be the last word on anything with legal or financial weight. Build the escalation path before launch, such as a text to a live rep or a flagged conversation in the CRM. Don't wait for the first bad answer.
Do referral and word-of-mouth leads need the same 90-second response?
They're less urgent, but don't skip them. A referred lead already trusts you more and is less likely to be calling three other companies at once, so a slower response costs you less. It still shouldn't sit in a queue until Wednesday. Same-day is a reasonable target for referrals, even without the pressure that comes with a paid ad lead.