You've Survived Every Wave, But This One Is Different

I had a math professor in college who turned from the blackboard where he was drawing a graph to face the class and say, “Those who understand compound interest collect it. Those who do not pay it.” Then he went back to calculating the slope of a changing curve at some particular point in time. The compound interest of a changing curve comes to mind when I think about what I experienced at Harvard’s Joint Center for Housing Studies' Remodeling Futures Steering Committee meeting at the end of April.

Dean Curtis, the CEO of Ingage, an AI sales presentation company, opened the day of AI presentations by describing a design-build firm somewhere in Texas that is down 20% in revenue. The loss didn't come from a recession, a bad project, or a key employee walking out the door. It came from a customer who took the designs they'd been given, ran them through an AI tool, found something marginally better, and went somewhere else. No confrontation, no warning. A quiet exit that the firm probably still doesn't fully understand.

Curtis’ story is a warning about complacency and about the danger of assuming you have more time than you do. He has spent thirty years watching technology reshape industries, from mobile databases in Army field packs to helping the iPhone unseat BlackBerry in under two years. He knows what a technology shift looks like before most people feel it.

The math has changed on waiting out AI

Every prior technology wave came with a grace period. Internet, email, smartphones, CRM software — first movers got an edge, but late movers could close the gap. Conservative business professionals could afford to let others make the mistakes, watch what works, then catch up. For many remodelers in many markets, that was a perfectly rational strategy.

But AI breaks that strategy in three ways at once. Your customers are leaving Google—they are asking ChatGPT, Claude, Gemini, and others to recommend remodelers in the area, and if you are not showing up in the answers, you do not exist to a growing share of potential customers. That share gets larger every month. Meanwhile, the competitor who started using AI eighteen months ago has an AI that has been learning about their specific customers, objections, and jobs. The AI's failures have sped it up, not slowed it down, because it is learning. The AI robot is smarter than anything you would start with today, and it gets smarter every day. Curtis broadened the opening example to include the remodeler’s competitor: another $12 million home remodeler, the one that adopted AI eighteen months ago, is now closing at a rate nineteen points higher than the laggard, and that gap compounds daily. The losing remodeler’s addressable customer base shrinks as more consumers turn to AI-assisted search, while their competitor's conversion rate grows as AI matures. The second remodeler is losing from both ends simultaneously.

50-Year-old company replaced call center employees with AI, and their customers prefer it

Alec Newcomb, Chief Marketing Officer at Thompson Creek Window Company, gave the room a view from inside. Thompson Creek is a 50-year-old company with a traditional sales model: in-home consultations, high-consideration purchases, and a customer base that skews older. The company now runs its call center with five people, down from fifty. The call center operates 24/7; its AI handles inbound calls, qualifies leads, and sets appointments.

Perhaps most surprisingly, 72-year-old customers who are given the choice between a human and an AI choose the AI. "If you had told me that before we tested it," Newcomb said, "I would not have believed you." That finding matters beyond the call center statistics. It challenges the assumption, common in remodeling, that customers want human contact above all else. What they want is competence, availability, and responsiveness. Newcomb's advice: test your assumptions before you bet on them.

Thompson Creek's results, 20 percent net sales growth, record marketing efficiency, and record margins, came from a deliberate organizational investment: hiring data scientists and software engineers, deploying agile practices, and building infrastructure in-house rather than buying from vendors who are not passing on the dramatic cost reductions in the underlying technology. Newcomb noted that the cost of running AI models has dropped precipitously since 2022, and almost no vendor has reduced its bills accordingly. Thompson Creek responded by building more of what they need themselves.

Your peers are already at the kitchen table with AI

The panel discussion that closed the session offered candor from remodeling leaders who are in the middle of figuring this out. Paul DesRoches, CEO of Moss Building and Design in Northern Virginia, described Natalie, a custom AI assistant trained on 25 years and roughly 7,000 of Moss's completed projects. The original intent was to give prospective customers a transparent window into the company's work: ask about project scope, get real examples with realistic price ranges. What happened was a surprise. Existing customers started using Natalie to ask design questions about their own past projects. Employees, including salespeople, started using her to validate their own estimates, pasting full contract scopes in and asking her to price-check their work.

"There is not a single person in this company, myself included, who knows all the details of 7,000 projects going back 25 years," DesRoches said. "Natalie does."

Mark McClannahan, CEO of Mosby Building Arts in St. Louis, built a business advisor first: a custom AI model loaded with the company's strategic identity, values, processes, and market positioning that his entire management team of twelve uses as a thinking partner. On top of that sits a layer of role-specific sidekicks. The sales sidekick knows Mosby's nine-step sales process, understands their products and pricing, and can prepare a salesperson for a call, coach a response to a difficult buyer, or review a proposal and flag weaknesses. Above that layer come specialists: AI agents that do work, not just advise.

Michael Anschel, principal at OA Design + Build + Architecture in Minneapolis, described a simple philosophy: find the toil, build the robot, and get the toil out of your people's way. His firm built a tool that cross-references construction drawing sets against project specifications, flagging conflicts a fatigued PM might miss after months on the same project. They built another that helps field staff identify which sealants and adhesives are compatible with which substrates, so carpenters don't need a chemistry degree to grab the right tube or roll of tape from the truck. Small tools, real problems, measurable relief.

AI handles the toil so your people can do the people work

Every speaker made the same point: this is about freeing people to do the work only humans can do. Judgment, relationships, and insight from lived experience get crowded out when talented people spend their days on repetitive work that a well-built AI could handle. 

Anschel named a risk worth taking seriously. AI defaults to the middle. It produces acceptable answers that look polished and authoritative. The companies getting the most from these tools learn to push back, asking AI to be critical, to argue the opposite, to surface what it doesn't know. That requires judgment; no prompt will replace. After the Chatbot writes your sales proposal, ask AI to critique it. After it writes code for a custom chatbot like Natalie, tell it to find the bugs.

Sean Corriel, Director of Product Management at Houzz, added the consumer side. AI-powered visualization tools are already changing what customers bring to the first meeting. Homeowners can photograph their kitchen, describe what they want, and receive a realistic rendering inside Google before they ever call you. The customer arriving at your door has already been designing. The remodelers meeting them with equivalent capability are winning that conversation.

Nobody in that room started with a master plan

Curtis closed with a practical challenge. Assess your team: find your explorers, your practitioners, and your builders, the people who can stitch AI into actual workflows and produce real outcomes. Identify the twenty percent of each role that only a human can do, then find someone who can build, whether that's a hire, a contractor, or an internal champion you develop. Newcomb's version: appoint champions inside each function, get your data in order, and experiment relentlessly.

The transparency AI enables with customers is a trust asset. Don't wait for the perfect policy. You will have today's policy, which you will improve tomorrow. None of the remodelers on that stage began with a master plan. DesRoches deleted a production database in his first ambitious attempt with Claude Code and kept going. McClannahan was frustrated before finding a framework that worked for his leadership team. Anschel started by building tools to solve his own daily annoyances and ended up with a library of them.

Someone in the audience asked the inevitable question: Are these panelists just exceptional early adopters, and how much of this is really relevant to the average remodeling company? No word on if this was the remodeler from the opening paragraph, but the honest answer is that it doesn't much matter. The technology is already inside the tools your customers use to research, visualize, and evaluate remodelers before they call you. Your potential customers are already asking chatbots who to hire, and you may be losing them without understanding why.

This article is based on presentations and panel discussions from the Harvard Joint Center for Housing Studies' Remodeling Futures Program. Speakers included Dean Curtis, Alec Newcomb (Thompson Creek Window Company), Sean Corriel (Houzz), Michael Anschel (OA Design+Build+Architecture), Paul DesRoches (Moss Building and Design), and Mark McClannahan (Mosby Building Arts). The session was moderated by Mark Richardson.