Where Automation Ends and the Human Touch Begins

By Pam Torrey, Senior Director of Marketing at Leaping AI 

Panda Exteriors started in 2019 as a power washing company. The three founders quickly realized one roof job could bring in the revenue of ten power washing jobs and pivoted into roofing, then spent a stretch as one of Tesla's largest subcontractors handling tear-off and dry-in for solar roof installations up and down the East Coast. That work spread crews and project managers across a wide footprint, infrastructure that made it easy to copy the model into residential roofing, siding, and solar. Today Panda is live in six states, Maryland to North Carolina, and still runs on the same engine it started with: canvassers knocking on doors. For the second episode of Built for Growth, I sat down with Christian Curry, President of Panda Exteriors, to talk about that business and where he decided technology belonged in it. 

Canvassing has had an uneven few years in home improvement. Companies spent much of the last five or six years backing away from it, pouring budget into Facebook ads, TikTok, and other digital channels instead. Panda never made that move, and Christian sees the industry drifting back toward door-to-door for reasons that go beyond nostalgia. “You're going out and earning the lead,” he told me. “You're not waiting for somebody to have the intent to search something online and click a link.” That control extends to cost too: once a canvassing operation reaches scale, a company sets its own cost per lead instead of renting leads from Angie's List or similar aggregators. And there's a second-order benefit Christian says matters even more than the leads themselves. Canvassing can create “a really, really good talent funnel,” as Christian put it. A person who is good at generating leads is easier to train into a sales rep than someone who has never worked with customers before.  

Where the Canvassing Handoff Breaks Down 

Behind every knock that turns into a booked appointment is a handoff. A canvasser earns a homeowner's attention at the door, then calls into Panda's call center, where a CSR verifies the homeowner's information, resets expectations, checks rep availability, and books the appointment into the CRM. It's a simple flow until you look at the shape of the demand hitting it. Canvassing crews hit the field around noon, but most people don't answer a knock until after five, when they're actually at home. That leaves CSRs sitting at a quiet desk from noon to five, then absorbing a hard spike from five to eight—a pattern that strains speed to lead just when an interested homeowner is standing on their porch waiting to book their appointment. 

The handoff carries a built-in tension too. “There's some friction built into it, and that's the balance of powers,” Christian said. A canvasser who hasn't talked to a single homeowner in five hours and needs to hit a quota will push almost any lead through. The CSR's job is to confirm it is qualified before booking the appointment. “You can imagine a canvasser building some form of resentment towards an individual that always kills my leads,” he told me, “when both of them are doing their jobs correctly.” 

Where Christian Draws the Line Between AI and Trust 

When Christian started looking at where automation belonged in that process, the line was less about the technology's limits than about what he didn't want it doing. “AI can handle the logistics, but...we wouldn't want it to handle the trust building aspect of it,” he said. The human at the door builds the trust and the urgency. The AI picks up everything after: confirming the information the homeowner already gave, writing it to the CRM, and routing the appointment to the right available rep. 

The hesitation about bringing AI in, Christian said, was mostly his own. “I was really nervous about how the homeowners would interact with the AI, how they would react to being put on the phone with an AI agent,” he told me. Panda piloted the change with its best rep first, someone with enough at-bats to absorb a bad call here and there, then used the early feedback to adjust. The first version of the voice agent asked homeowners to repeat the information they'd just given the canvasser at the door, and homeowners pushed back almost immediately: why am I saying this again? Panda changed the script from asking for details to confirming them, reading back information instead of requesting it fresh. “It turns out homeowners aren't hesitant to talk to an AI,” Christian said. “They just want it to be quick and easy.” 

The Math Behind Automating the Confirmation Call 

Christian's approach to deciding what to automate comes from EOS, the operating system Panda runs the business on, specifically a concept called delegate and elevate. It sorts a person's responsibilities into quadrants, one of which is the tasks they're bad at or simply dislike. “If you can find high value individuals that have high value skills and take the things that they hate doing, typically a repetitive, monotonous task, and figure out a way to automate that,” he said, “that gives them the time back to do those high-level things.” 

Running that lens over the confirmation call turned up a number. Panda's canvassing team alone carried a customer acquisition cost hovering around 10 percent. Layering in everything happening on the scheduling and confirmation side added another 3 to 5 percent on top. Automating those calls let Panda claw back that 3 to 5 percent, and just as important, freed the five CSRs who'd been handling it to move into outbound telemarketing, generating their own leads instead of supporting leads that already existed. The AI side of the ledger helped too: no sick days, no attrition, no ramp-up when opening a new territory. “It's very easy copy and paste from location to location for us,” Christian said. Acquisition costs dropped, and revenue from the redeployed team's outbound calling climbed at the same time. 

Two things surprised Christian once the change was live. The first was how quickly the right team members leaned into the shift, at a moment when the industry's dominant story about AI is job loss. “Seeing the ones that were the right people, seeing them just want to elevate and move to higher level tasks, was pretty surprising,” he said. The second was how little resistance came from homeowners once the script was fixed. “They're at their door. They want to get back to their day,” he said. “Whatever's easiest for me to get this appointment booked, head back to my life.” 

Asked what he'd tell another home improvement leader eyeing the same move, Christian's advice starts with the back office, not the tech. Look for the repetitive, copy-paste tasks first, the ones where someone is manually moving information from one place to another. And leave the front line alone. Panda's canvassers didn't change a device, an app, or a workflow; the only difference was a different number to call. “The less disruptive you can be to the people in the front, the better off you're going to be getting the user adoption that's needed,” he said. His test for knowing if a business has gone too far in the other direction is simple: if a customer wants to talk to a human and can't easily get to one, something's been over-automated. 

Listen to the full conversation with Christian Curry on Built for Growth, Episode 2, including his rule of thumb for knowing when a business has over-automated and how compliance requirements shift from state to state.