Tony Marchetti had been running Marchetti Construction Group for 18 years, and he could feel the industry changing under his feet. His company had built a solid reputation in Bay Area commercial tenant improvements and ground-up retail — $65 million in annual revenue, 140 employees, and a client list that kept coming back. But the margins were getting thinner, and Tony knew exactly why: coordination overhead.
Every project generated thousands of documents. RFIs, submittals, change orders, daily logs, inspection reports, safety documentation. His project managers were drowning in paper — or more accurately, drowning in email. Tony estimated that his PMs spent 40% of their time just managing information flow rather than managing construction. And when things fell through the cracks, as they inevitably did, the cost showed up as delays, rework, and change orders.
The breaking point came on a $12 million retail build-out in San Jose. A subcontractor's question about a mechanical specification had been buried in an email thread 47 messages deep. By the time the PM found it, two weeks had passed. The answer required a design revision. The revision triggered three related change orders. Total cost: $180,000 and a 16-day schedule slip. Tony was furious — not at the PM, who was doing the best anyone could with the tools available, but at the system that allowed a simple question to become a six-figure problem.
Tony's estimator, Linda Vasquez, had connected with Hamilton-Blackwell at a Procore user group meeting. She brought back a description of their construction-specific discovery process that intrigued Tony. "They don't just look at your software," Linda told him. "They watch how your PMs actually communicate with subs. They sit in trailers and watch people work."
That's exactly what happened. The HB team spent a week rotating between Marchetti's office and two active job sites. They sat in project meetings. They watched PMs process RFIs. They tracked every communication between the general contractor and its subcontractors on a $9 million project in real time. By Friday, they had identified what they called the "information decay chain" — a systematic pattern where critical information entered the system via one channel, was needed in another, and degraded as it passed through human intermediaries.
The solution targeted the decay chain directly. Hamilton-Blackwell deployed an AI layer on top of Procore that automatically classified every incoming communication — RFI, submittal, change order request, information request, schedule question — and routed it to the right person with the right context. The system tracked response deadlines and escalated automatically. It summarized long email chains into actionable bullet points. And critically, it flagged communications that contained embedded questions or decisions that might otherwise be overlooked.
For estimating, HB built an AI tool that analyzed Marchetti's historical bid data — 12 years and over 400 projects — to identify pricing patterns, cost escalation trends, and risk factors specific to their market and trade mix. Linda could produce a preliminary estimate in a third of the time, with the AI flagging line items where historical data suggested the estimate might be low.
The results on the first fully AI-augmented project were dramatic. A $14 million mixed-use project in downtown Oakland completed with zero preventable change orders — compared to Marchetti's historical average of $320,000 in preventable overruns per project of that size. RFI response time dropped from an average of 8.3 days to 2.1 days. Daily reports, which superintendents previously spent 30–45 minutes compiling, were being generated in under 5 minutes from field data the supers were already entering.
Tony did the math at the end of the first quarter. Across three active projects, the AI systems had eliminated an estimated $340,000 in costs that would have previously been absorbed as overhead, rework, or margin erosion. His PMs were working fewer hours and producing better outcomes. And his superintendent retention — always a challenge in the Bay Area market — had stabilized for the first time in years.
"I'm a builder," Tony said at a subcontractor appreciation dinner. "I'm not a technology guy. But I know what it feels like when a project runs clean, and I know what it costs when it doesn't. This is the first time I've seen technology that actually makes the job site better instead of just adding another screen to look at."