AI moves from pilot to platform across global construction operations

Nearly half of the 157 companies listed in South Korea's Korea Proptech Forum 2026 member directory have adopted AI as a core business function, according to Chosun Ilbo. That share, drawn from a sector that until recently treated technology as optional overhead, marks a clear inflection point. The forum published its first dedicated 'AI Edition' this year, mapping 75 AI-enabled companies across five functional units: development, construction, transactions, marketing, and asset management.

The broader construction AI market is on pace to grow 24.7% annually, Chosun Ilbo reported. That figure reflects not just startup formation but active deployment by some of the largest contractors in the world, each targeting a different operational bottleneck.

From job site data to enterprise feedback loops

DL E&C is the most structurally ambitious of the Korean deployments. The firm was the only domestic construction company to present an AI-driven innovation case at Palantir Technologies' APAC Summit Korea 2026 in Incheon in April, according to Chosun Ilbo. DL E&C first adopted Palantir's Foundry platform in 2022 and has since built what it calls a 'Flywheel ecosystem,' connecting design, construction, and maintenance data into a single decision layer. More than 87 years of accumulated data covering costs, quality, safety, and design now feed live project planning and decision-making meetings. Work instructions generated during construction automatically populate planning records, so past change orders and risk events surface proactively when new projects are scoped.

GS Engineering & Construction took a different approach, developing its AI Defect Prevention Platform internally and integrating it directly into its quality management system. The platform analyzes defect types and causes by construction process, visualizes cases in 3D, and is designed to be readable by foreign workers on multilingual job sites. GS E&C reported zero defect judgments in two consecutive reviews by South Korea's Ministry of Land, Infrastructure and Transport over the past year, as Chosun Ilbo noted.

Daewoo Engineering & Construction's deployment targets a different risk: AI hallucination in contract and specification review. Its Baro-Dap AI is a vertical model trained exclusively on internal contracts and specifications, designed to answer only from within that document corpus. A companion tool, Baro-Letter AI, handles document drafting. By grounding the model in proprietary data rather than general web training, Daewoo is addressing a known reliability concern that has slowed enterprise adoption of generative AI in legal and compliance-sensitive workflows.

Samsung C&T is collaborating with AWS on an AI agent for construction operations, and has also deployed an automated steel bolt-tightening robot for structural fastening work at height, incorporating automatic bolt-position recognition and error correction to standardize output quality, according to Chosun Ilbo. Hyundai Engineering & Construction operates a generative AI-based sales consultation service on the transaction side.

Supply chain failures are the real cost driver

While Korean firms are advancing AI at the job site level, a parallel challenge is being addressed on the procurement side. Forbes contributor Sabbir Rangwala pointed to megaproject case studies including Hudson Yards in New York City, the Big Dig in Boston, and the Burj Khalifa in Dubai as examples of the cost and schedule overruns that plague construction at scale. Hudson Yards specifically was affected by a shortage of steel and other building materials tied to market demand, tariffs, and production constraints, contributing to billions of dollars in overruns, according to Forbes.

The structural problem, as Forbes described it, is that construction has not historically adopted the design-for-manufacturing and design-for-reliability disciplines that semiconductor and other precision manufacturing industries use to account for supply chain constraints early in the design phase. Every construction project is effectively a one-off, making it harder to encode prior knowledge into future planning without AI-assisted tools.

Krane, a startup founded in 2022 by CEO Eshan Jayamane, is targeting exactly that gap. Jayamane, who has a background in large-scale industrial, energy infrastructure, data center, and hospital construction, built Krane to provide a unified real-time platform managing submittals, lead times, purchase orders, and delivery schedules across thousands of suppliers and line items, according to Forbes. The company recently raised $9 million to expand those capabilities, and its platform is designed to surface supply chain risks during the design phase, when projects can still absorb changes without triggering costly scope pivots.

What this means for your team

• Evaluate vendor AI for hallucination controls: Daewoo's approach of grounding AI responses solely in internal documents is a practical model for any procurement or legal team considering AI for contract review. Ask vendors how their system handles out-of-corpus queries before deployment.
• Assess whether your project data is structured enough to feed a platform: DL E&C's Foundry deployment draws on 87-plus years of cost, quality, safety, and design records. If your historical project data is siloed or unstructured, that is the prerequisite problem to solve before AI can deliver feedback-loop value.
• Build supply chain visibility into design-phase workflows, not just construction: Krane's model explicitly targets the design stage, when lead-time and procurement constraints can still be designed around. Project managers running active procurements on large builds should be evaluating real-time supply chain platforms before steel orders are placed.
• Use the Korea Proptech Forum's AI Map as a competitive benchmark: With nearly half of 157 member firms now running AI as a core function, the map provides a useful signal of where the industry's baseline is moving and which functional gaps remain underserved.