DiscoverConstruction Technology Podcast
Construction Technology Podcast
Claim Ownership

Construction Technology Podcast

Author: Luigi La Corte, Mike Powers, Andrew Zukoski

Subscribed: 0Played: 1
Share

Description

Three venture-backed construction technology CEOs discussing and debating the most important changes and trends in the construction industry.

36 Episodes
Reverse
Chris Boyd, Chief Product Officer at Trunk Tools, joins Luigi, Mike, and Andrew to open up the numbers behind how AI reads construction drawings, walking through per-object F1 benchmarks and cost-per-sheet comparisons between Trunk Tools' internal "Cortex" models and the latest frontier models. He explains why the platform is modular—roughly 21 models hot-swapped by job to be done—and how object detection powers fully automated drawing-revision comparison that catches unclouded changes, including one owner directive that would have triggered a roughly million-dollar change order. The conversation covers building a knowledge graph across drawings, specs, and RFIs, flat project-based pricing so field teams never weigh token costs, and measuring ROI on operational metrics like submittal cycle time rather than hours saved. Chris also shares his view on AI-native applications as orchestrators across project systems, the construction labor and knowledge-transfer gap, and what winning looks like for field teams.
Geoff Lewis, SVP of Product at Procore, joins Luigi, Mike, and Andrew fresh off Groundbreak to unpack Procore’s “see, understand, act” vision with DroneDeploy and Datagrid, and why customers still cheer hardest for core wins like bulk-editing submittal workflows. He maps the next billion-dollar growth vectors—owners (including a new capital planning product), trade contractors, and international—while defending a customer-centric product culture that stuck with him from Honest Buildings through Procore’s climb from ~$400M to ~$1.6B in revenue. The conversation digs into Procore AI packaging (starter-pack agents vs. pro/enterprise credits), embedding agents into everyday workflows like bid leveling and RFIs, and how token economics and “digital coworkers” change pricing and adoption. Geoff also weighs open-weight vs. frontier models, physical AI and robotics on site, and why competing head-on with an AI-accelerated Procore remains a tough slog for AI-native CMS startups.
Clifton Harness of TestFit joins Luigi La Corte, Mike Powers, and Andrew Zukoski for a deep dive on the Model Context Protocol (MCP). Andrew lays out the vocabulary of MCP servers, clients, harnesses, models, skills, and tools, and the group debates when MCP beats direct API access, from latency for high-volume calculation engines to permission scoping and IT governance. Clifton shares that TestFit's internal usage grew roughly 30x after it released its MCP server in August, and the hosts argue over whether agents and natural-language interfaces will make PMIS user interfaces like Procore's less necessary. The conversation closes on structured construction documents, "AEC compilers," performance-based specs with dependency graphs, and whether AI could ever design a perfect building.
Josh Whisenant of Roger joins Luigi La Corte, Mike Powers, and Andrew Zukoski the day after Procore's Groundbreak conference in Orlando. The group recaps Procore's AI announcements, including the Magpie model and agent "starter packs," debates whether agents are really different from skills, and discusses why contractors should keep benefiting from foundation models that get better and cheaper every few weeks. Josh walks through Roger's relaunch as a services provider that carries general contractors to operational outcomes, covering buyout and scope writing, submittal review, and a billing approach where Roger doesn't send an invoice if the work isn't a fit and puts its fee at risk for scope gaps. The hosts compare who actually owns scoping and buyout at GCs (precon, procurement, or project management), and Josh shares his thesis that AI can industrialize the contractor and change, rather than eliminate, operations roles. He closes with where Roger fits best: complex, fast-moving vertical projects and high-volume submittal work such as data centers.
Luigi La Corte, Mike Powers, and Andrew Zukoski dig into how far general-purpose AI models can go on construction work now that frontier labs are buying expert-labeled, proprietary data, and where project data, harnesses, and purpose-built UI still matter. They compare AI adoption across mid-market GCs, mechanical reps, and manufacturers, from enterprise ChatGPT and Claude licenses to shared skills, MCP connectors, and the low-hanging fruit in everyday document workflows, plus whether bid management goes "headless" inside the inbox. Mike breaks down LG's long-term agreement with mechanical rep group Air Controlled Concepts, which backs a new 350,000-square-foot US chiller factory and more than five gigawatts of capacity, and why aggregated rep demand helps manufacturers commit to building capacity. The episode closes with a walkthrough of a closed-loop data center cooling design using dry coolers and chillers, and what it means for the debate over data center water use.
loading
Comments