July 4, 2026 • By Frank DeLuca

Your Builder's Project Manager Left on Day 147. The Replacement Has Never Seen Your Foundation.

On a Wednesday in March, Michael Torres walked into his superintendent's office trailer at 6:40 a.m., set his hard hat on the plywood desk, and told the owner of a $1.2 million custom home in suburban Denver that he'd accepted a position with a national builder. His last day was Friday. He'd been running their project since the permits cleared in October.

By the following Monday, a new project manager named Kevin was standing in the framed-out kitchen, holding a clipboard he hadn't written on yet, looking at a chase between the second floor and the garage where the HVAC ductwork and the electrical panel were supposed to coexist. Kevin had been with the company for eight months and had never managed a project this size alone. His first question was straightforward: "Where's the mechanical plan?" His second was less so: "Did anyone tell the HVAC sub he could start rough-in early?"

Nobody had told Kevin, because nobody had written it down. Michael had texted the HVAC contractor directly from his personal phone on January 14 to say go ahead and start ductwork in the master suite wing a week early because the framers were ahead. That text thread, along with every schedule adjustment, substitution decision, and sub negotiation it contained, left the project when Michael did.

The Math Nobody Runs

Bridgit's 2026 Construction Workforce Benchmark Report, which analyzed data from 233 companies and 114,000 workers including nearly 40% of the ENR 400, puts the non-senior project manager attrition rate at 14.3% annually. For superintendents, it's 15.4%, but the senior versions of those roles barely move, with senior PM attrition at just 3.6% and senior superintendent attrition at 4.1%. Whoever is running your custom home is almost certainly not the senior version.

A custom home takes 10 to 14 months from permit to certificate of occupancy, and running 14.3% annual attrition across a 12-month build, the probability that your project manager leaves during your project is roughly one in seven. That's not a rounding error you can ignore, and it's not a tail risk you can insure against. It's a structural feature of the industry.

For a builder running 10 simultaneous projects, the probability that at least one PM departure hits a project in the next year approaches 78%. This isn't bad luck. It's Tuesday.

What Actually Walks Out the Door

I've managed projects for twenty-two years. I've been the guy who left and I've been the guy who replaced the guy who left. The formal documentation, the stuff that lives in the project folder, the plans, the permits, the specs, the change orders with signatures, that survives and accounts for maybe 60% of what you need to run the project, but the other 40% lives in places no filing system reaches.

Consider what that 40% actually contains: a text thread with the plumber where you worked out that he'd rough-in the guest bath on Thursday instead of Friday because his crew had a conflict. A verbal agreement with the homeowner that the kitchen island would shift 14 inches east to avoid the floor drain, discussed at the site visit but never formalized because it didn't affect the price. Your four-month read on which sub shows up when he says he will and which one needs a confirmation call at 5 a.m. A quarter-inch proud bottom plate on the east wall that the framing crew left and the drywall crew will need to shim around, something you noticed during a walkthrough and filed in your head because it wasn't worth a formal note.

All of it. Gone in three business days.

The Reconstruction Period

Industry terminology calls it "knowledge transfer" and treats it like a checklist you can execute over a lunch meeting, but in practice it's two to four weeks of the new PM re-learning a project that someone else built the institutional memory for. They walk the site with fresh eyes, which sounds like a benefit until you realize that fresh eyes don't know the structural engineer changed the header spec on April 3 or that the homeowner rejected the first window sample because the bronze anodizing ran too warm.

During the reconstruction period, schedule decisions stall and sub coordination lapses because the new PM doesn't have the relationships yet, and field questions that the old PM would have answered in a text take three days because the new PM needs to read the spec book first. A study compiled by OpenSpace found that construction teams lose $280,000 annually to coordination challenges even when the team is stable, and adding a personnel change into that equation means the coordination loss climbs in ways that are difficult to isolate on a line-item budget.

For the homeowner on a $500,000 custom build carrying $3,000 to $5,000 a month in construction loan interest, two to three weeks of effective delay during reconstruction costs $1,500 to $3,750 in direct carrying costs alone. That doesn't count the rework risk from uninformed decisions, the schedule compression that inevitably follows as teams scramble to make up lost ground, or the overtime the builder will eventually eat to get back on track.

CII found that a 10% increase in turnover results in a 2.5% increase in total project labor costs, which on a $500,000 house where labor runs 40% of hard costs translates to an additional $5,000 absorbed somewhere in the budget. It never shows up on your change order log; it shows up in the builder's margin, which eventually shows up in the quality of the last 10% of your house when the budget gets tight.

Where the Documentation Actually Lives

I surveyed three custom home builders I've worked with over the past decade about where their project knowledge resides, and a consistent pattern emerged. Formal project management software (Buildertrend, CoConstruct, or a spreadsheet that pretends to be project management software) holds the schedule, the budget, and the permit documents. That's the 60%.

The rest breaks down roughly like this: text messages on the PM's personal phone account for about 30% of day-to-day coordination. Email, usually the PM's company email but sometimes personal, carries another 20%, while verbal agreements, site conversations, and phone calls that never get memorialized in writing cover the remaining 50% of the informal knowledge, though these numbers overlap because some topics get discussed in multiple channels without any of them producing a document of record.

When the PM leaves, the text threads on their personal phone go with them, company email stays but nobody reads 800 messages to reconstruct context, and verbal agreements are simply gone. The homeowner has their own text thread with the PM, which gives them a partial record, but they don't know what conversations happened between the PM and the subs that never included them.

One builder told me his PM left and the replacement discovered three weeks later that the homeowner had approved a tile upgrade in the master bath during a site walkthrough. No change order. No email. "I'll take care of the paperwork," he'd said, and then gave his two-week notice before he did, by which point the tile had already been ordered. The homeowner insisted the upgrade was discussed and approved at the original price. The builder ate $2,800.

What AI Project Memory Could Do (And What It Can't)

The commercial construction world has been building toward "project memory" tools for a decade. Procore holds 2.4 million projects on its platform, OpenSpace captures 360-degree imagery that creates a visual record of every corner of a job site on every visit, and Buildots uses fixed cameras and AI to compare as-built conditions against BIM models in real time. Sitemetric just launched what it calls "agentic AI" that monitors construction sites continuously, answering natural-language queries about who was on site, which subs are missing credentials, and how labor hours are trending.

None of these tools were designed for the $500,000 custom home market, because their pricing, complexity, and implementation requirements assume a project with a dedicated office trailer, a full-time PM, and a budget that can absorb a $3,000-to-$5,000 monthly software subscription. A custom home builder running five projects with two PMs and a bookkeeper is not their customer, and the builder knows it.

The gap is in the middle: what a residential custom home needs is not a construction management platform. It's something simpler: an AI that reads the PM's text messages, emails, and daily notes, extracts the decisions, commitments, and open items, and maintains a running "state of the project" document that any replacement PM could pick up and understand in an afternoon instead of three weeks.

LLMs can parse conversational text, identify commitments ("I told the electrician to start panel work Monday"), flag unresolved items ("homeowner asked about cabinet hardware, no response logged"), and generate structured summaries. Implementation cost would be next to nothing compared to what reconstruction actually runs.

But three problems stand between the concept and the product. First, privacy: a tool that reads a PM's text messages raises legitimate questions about surveillance, especially on personal devices. Second, adoption: the PM who needs it most is the PM who is least likely to install it, because the kind of PM who texts subs from a personal phone and doesn't write change orders is also the kind of PM who won't adopt a documentation tool. Third, the people who would benefit most, homeowners, have no visibility into the problem until it's too late.

What You Can Actually Do

If you're building a custom home, you can't control your builder's PM attrition rate, but you can control how much of your project's knowledge survives a departure.

Insist on a shared communication channel. Not text messages between you and the PM that disappear the moment someone changes jobs. A platform, even a free one like a shared Google Drive folder or a Slack channel, where decisions, approvals, and schedule changes get documented in a place the builder controls and the PM can't take with them. This isn't about the technology; it's about establishing that project knowledge belongs to the project, not the person.

Get every change in writing before work proceeds. Not "I'll send you the paperwork later," and not a handshake on the site visit that nobody memorializes. The paperwork comes first. If your PM discusses a change at a site visit, follow up that day with an email: "Confirming today's conversation that the island is moving 14 inches east at no additional cost. Please confirm." If they don't confirm, the change doesn't happen.

Ask your builder directly what happens to your project if the PM leaves, and if they don't have an answer, that tells you something important about how they think about continuity and whether they've planned for the turnover that their own industry data guarantees.

Builders who have thought about it will describe a handoff process, documentation standards, and overlap between the departing PM and their replacement; builders who haven't will tell you it won't happen.

It happens to about one in seven.

Limitations

Bridgit's benchmark covers all construction, including commercial, so the 14.3% attrition rate may not perfectly map to residential custom homes, where PM turnover may run higher (smaller firms, lower pay, fewer advancement paths) or lower (more personal relationships, less bureaucratic). No study has isolated the mid-project departure rate for residential PMs specifically, and my knowledge reconstruction cost estimate is based on carrying costs and typical delay timelines rather than a controlled study. The informal knowledge breakdown (text vs. email vs. verbal) reflects conversations with three builders, not a representative sample.

That 78% probability that at least one PM departure hits a 10-project portfolio in a year assumes independence between departures, which isn't realistic. PM departures tend to cluster when a builder loses a key leader or when a competitor enters the market with higher pay, so the actual probability could be higher or lower depending on the builder's specific retention practices and local market conditions.

Frank DeLuca has spent twenty-two years managing construction projects, watching the important details disappear into text threads, and writing about the operational gaps that technology hasn't closed yet.