Cutting a Nine-Month Process to Thirty Days
Joe Johnson works in AI, but he’d rather talk about fire trucks, commercial refrigerators and surveyor PDFs: the specific, tedious problem of getting what experienced people know out of their heads and into software. He’s been a VP of Cloud Operations and a Director of IT, and he’s run M&A integrations. Today he’s building the software and data team at Perry Homes as VP of AI & Customer Experience Technology. Perry Homes is one of the largest private homebuilders in Texas, in business for 50+ years, building roughly 4,500 homes a year across Texas and Florida. Before that he was at Revalize, a commercial SaaS company that sells to complex manufacturers, and that’s where his most interesting AI work happened.
You can find Joe on LinkedIn and on his website.
$50k of Software, But It Needed $200k Of Services To Set Up
Revalize sells CPQ (Configure, Price, Quote) and PLM software to manufacturers of complicated products: commercial kitchen equipment, fire trucks, specialized industrial machines. A single unit can have 1,200 configurable options, and some combinations just don’t work. A two-inch flange can’t go into a four-inch pipe on a fire truck. A left-opening door won’t fit a right-opening fridge.
The software could handle all that. Getting customers set up on it was the hard part.
If you bought $50,000 of software from Revalize, you’d often need $200,000–$300,000 of consulting to implement it properly, because of all those small rules. Implementation meant sitting down with the customer’s most experienced salespeople and getting everything they knew out of them: every option combination, every conflict, every odd case. It took six to nine months, and it was the biggest thing standing between a signed deal and a live customer paying money.
Most of that knowledge wasn’t written down anywhere. It was in the heads of salespeople who had quotas. “Am I gonna sit down and spend eight hours downloading my brain to these consultants from the software company,” Joe says, “or am I gonna go out and sell fire trucks? I’m gonna go out and sell fire trucks.”
The Solution: Give the LLM the PDFs
Joe’s team noticed something simple: all of this was already in the product PDFs. So they gave the PDFs to an LLM, had it build an Excel compatibility matrix of options (a checkbox for combinations that work, an X for ones that don’t), and sent it back to customers to check.
Nine months turned into roughly thirty days. “From an adoption standpoint, I can tell you, like, it was instant hockey stick,” Joe says. “Everybody started using it.”
The new limit was how fast sales could close deals. “The biggest problem we had now is we couldn’t sell fast enough to keep up with our ability,” he says. “Sales, go sell 500 more customers, because now we have the capacity.”
Even before a customer sent any data, the team could take public product information and reach 60–65% accuracy. That was enough to load a demo environment with the customer’s real catalog before the first meeting. “It blows the customer’s mind that we put that much effort into it,” Joe says, “because they’re thinking we sat someone down to click all the if-then statements in the app instead of just using AI to cheat it.”
But Then: The ROI Question
Joe keeps going past the happy part.
The goal was to cut a nine-month implementation to three months, prove it with a real customer and show the results. They did it: from kickoff call to the customer processing orders took three months and two days. “Close enough for government work,” he says.
Then came the harder question of what that actually saved. Revalize did four implementations a year. Proving the process had cost $70–80K of people’s time. “It didn’t really feel like a big payoff in the end,” Joe says, “because we picked something that sure seems bright red flashing emergency, let’s get rid of this, let’s solve this problem, but at the end of the day, the level of effort versus the return was not there.”
His advice to builders is to find the pain point and then put a number on it. “What is the pain point and the number associated with the pain point of your success? What is the happy path and in that number?” An impressive technical result on a problem that touches four customers a year is still the wrong thing to have worked on. “Is it worth $60,000 worth of developer time to save something that’s going to maybe save $5,000 a year?”
All through the project, Joe wanted to push past 65% pre-sales accuracy toward an implementation that needed almost no people. His colleagues wanted to call it finished. “You have to aim for the 10X goal, knowing you’re gonna miss and hit the 2X goal,” he says. “But you have to be audacious.”
There’s another piece of the ROI math. As model prices have gone up, the numbers have started to flip for companies that leaned hard on expensive frontier models for work that doesn’t need them. “It’s now cheaper for me to outsource development back to the Philippines than it is to pay my token budget to go do development using AI,” Joe says. His advice: pick models on purpose. “Maybe Sonnet 4.5 is good enough. We don’t need to use Opus for everything.”
The 96% Problem
Joe also has firm views on Copilot rollouts and what “adoption” really means.
What he saw: private equity backers or pushy CEOs wanted the company to “have AI,” so it bought Copilot. That made some sense. They already spent a lot with Microsoft, and a couple of years ago Copilot wasn’t that far behind ChatGPT or Claude. What they didn’t think about was lock-in: a three-year contract, Microsoft products only, and no easy way to switch as the models got better. “They later come to regret that,” Joe says.
A year later, 96% of employees use Copilot weekly, about 30 interactions per employee per week on average. Microsoft’s reports showed roughly $80,000/month in time saved.
“And that translates to $0 of savings. People aren’t more productive. Nor are they reducing headcount. Nothing really changed.”
He puts it this way: “96% of people use the bathrooms at the office every day. Does that mean that makes us a more valuable company because we have bathrooms? Maybe not having them would make us a less valuable company. But it’s meaningless.”
A platform that reports usage will always show usage. Whether the way the company works has changed is a separate question, and it’s the one that counts. Joe measures by whether workflows change.
The Dirt
Three and a half months into Joe’s time at Perry Homes, the clearest AI win is about ordering dirt.
Before any house goes up, the lot has to be level. Surveyors measure the grade across the whole plot and produce a PDF: a picture of the land with height differences written all over it. Someone types those numbers into a calculator to work out how much fill dirt to order. Get it wrong and either you can’t finish the pad or you’ve paid for a truckload you don’t need. Errors were running at about 5–6%.
Joe’s team used Azure Document Intelligence to pull the numbers out of the surveyor PDFs and feed them straight into the calculator. The error rate fell to 0.5%.
“People imagine robots building 3D-printed houses,” Joe says. “And it’s like, no, we just order dirt better than anybody else right now.”
Perry Homes closes 4,500 homes a year, and every construction delay costs money twice over: carrying costs while the company waits to close, and a real problem for a customer whose house isn’t ready when promised. A buyer expecting to close on July 15th who gets pushed to July 25th may have already ended a lease, booked movers and set up bridge financing. A mistake in the dirt order costs a lot more than one extra truckload.
“The most successful things,” Joe says, “are probably the least sexy of all of it.”
The Training Pyramid
Joe built an AI training program at Revalize and is building it again at Perry Homes. It has three levels: Foundation, Fluency and Mastery.
Foundation is required for everyone who uses a computer for their job: what AI is, how LLMs work, how to structure a prompt, how to tell when the model is lying to you. At Revalize the target was 100% completion across roughly 800 employees worldwide, about 20 hours of training in total, by May 31, 2026. Joe says his old team made it.
Fluency is for managers and people leaders: how to build AI into the team’s workflows and find where it actually helps.
Mastery is for very few people, around five at an 800-person company. They take outside graduate programs and AI certificate courses and become the in-house experts who make everyone else more effective.
The uneven shape is intentional. “You don’t need 800 AI masters,” Joe says. “You need 800 people who aren’t afraid of it, and five who are obsessed.”
He treats AI literacy the way he treats web search: “I trust people who know how to use a search engine. Now I need to trust people who know how to talk to AI and to do it efficiently.” Foundation gets construction workers in the field to where they can use the tools the five obsessives build for them.
Citizen Development at Perry Homes
Joe is also setting up what he calls “citizen development” at Perry Homes. He likes that term better than “vibe coding,” which he thinks makes the work sound less disciplined than it is.
Perry Homes’ product owners (their name for IT project managers, because they want people to own outcomes, not just carry out tasks) now use AI tools to build working UI prototypes together with business stakeholders. These are real, clickable screens, built live during the conversation.
“They don’t know what they don’t know,” Joe says of business stakeholders. It’s hard to get good product requirements from someone who has never built software, because they can’t say what they need when facing a blank page. Put something on screen, however rough and however obviously a prototype, and the feedback gets specific fast.
From there the requirements go into a structured pipeline. BSAs write PRDs, ERDs and API contracts formatted to serve as context for Cursor, the AI coding environment Joe’s team uses. Developers then feed that documentation into code generation, with far less back-and-forth between requirements and the build.
“We’re front-loading that investment of time into getting documentation and context really well architected,” Joe says, “to create high-quality software.”
It’s the same lesson as the Revalize work: what the AI produces depends almost entirely on what you give it. Joe’s team hands Cursor a fully structured specification, reviewed by people, before anyone writes a line of code.
Find Joe Johnson on LinkedIn. Learn more about Perry Homes at perryhomes.com.