Hanmi Global hosted the Global PM Summit 2026 in Seoul on Friday, where researchers and industry leaders demonstrated construction AI tools now entering commercial use. The technologies, which generate 3D building models from text commands and automate project risk analysis, arrive as the sector faces declining productivity, an aging workforce, and rising costs.
Text-driven BIM and autonomous agents move beyond the lab
André Borrmann, director of the Construction AI Center at the Technical University of Munich, presented multi-agent AI systems that create building information models from text prompts alone. In modeling tests, the AI achieved an 86 to 95 percent success rate on tasks such as wall generation and opening insertion. Borrmann also showed AI agents that can recognize and manipulate BIM software screens directly, without custom integrations.
Combining on-site photos with BIM data lifted the identification accuracy of structures like formwork by up to 50 percent, Borrmann said. Pre-training robotic arms on simulated construction tasks pushed their work success rate above 70 percent. The university signed a memorandum of understanding with Hanmi Global to pursue joint research on AI-based project management innovation.
Half of PM work could be automated, says Turner & Townsend COO
David Whysall, chief operating officer of Turner & Townsend, told attendees that a significant portion of project management work can be automated through AI - expanding to as much as half. "The core value of AI lies in strengthening the capabilities of professional personnel and advancing data-based decision-making," Whysall said. His firm is building a platform that standardizes and integrates project data globally, aiming to deliver services across cost, schedule, and risk management.
Whysall's keynote framed automation not as headcount reduction but as capability amplification for existing teams. The emphasis on data standardization reflects a broader push across the industry to make fragmented project information machine-readable before applying AI.
A realistic autonomy target: AI decides, humans supervise
Rafael Sacks, director of the National Building Research Institute at the Technion - Israel Institute of Technology, broke construction production management into four functions - information gathering, analysis, decision-making, and execution - and mapped them across six levels of automation. He argued that full autonomy is unlikely given the industry's safety and contractual obligations.
"Given the nature of the construction industry, where safety and contractual responsibility are important, an intermediate level of autonomy in which AI decides and executes while humans supervise and can intervene will be a realistic goal," Sacks said. The Technion institute also signed an MOU with Hanmi Global for joint technology trend surveys and market response strategies.
For managers tracking how AI is reshaping project delivery, tools like these are moving into the AI for Real Estate & Construction stack faster than many expected. The summit made clear that BIM, computer vision, and robotic execution are converging into workflows that teams can begin testing now, not years from now. Professionals looking to build practical skills can start with an AI Learning Path for Project Managers that covers risk analysis and workflow automation relevant to these developments.
Why this matters for management
The technologies shown in Seoul are not prototypes. They are hitting the commercialization stage with specific performance data - 86 to 95 percent modeling accuracy, 50 percent gains in structure identification, 70 percent robotic task success. For project managers and operations leaders, the immediate question is whether their data infrastructure is clean and standardized enough to feed these tools. The automation ceiling Whysall described - up to half of PM work - depends less on AI capability than on data readiness and team willingness to adopt supervised autonomy models like the one Sacks outlined.
Your membership also unlocks:
