← Cases
TaiwanConstruction and engineering2024 – ongoing

Kedge Construction

AI on the Guan De Wanhua construction site

Kedge Construction cut cost estimating from 3-4 months to 1-2 weeks with AI

WorkingCompany's own claim

Problem

Construction in Taiwan faces a labour shortage, and formwork workers are mostly over 50. Cost estimates took months, and architectural renderings were outsourced.

Approach

The company began planning AI use in 2024 and ran trials for about two years. On the Guan De Wanhua site it used a spray-painting robot whose height, thickness and distance are set with AI, AI cost estimation, BIM with AI for renderings, and AI image recognition that detects missing hard hats and sends alerts. Staff review the AI estimates. Workers still finish edges by hand.

Results

from 3-4 months to 1-2 weekstime to prepare a cost estimate, as stated by a company executiveCompany's own claim[1]
from 3-4 weeks to about 1 weektime to produce architectural renderings, now done in-house instead of by an outside firm, as stated by a company executiveCompany's own claim[1]

Takeaway

The gains came from specific tasks such as estimating, rendering and spraying. An executive said many subcontractors stayed conservative and that robots may not pay off at once for skilled trades such as painting and plastering.

Categories

Business function
Production and quality
Innovation type
Process innovation
Evidence
Company's own claim

Sources

  1. AI進入工地 根基營造「創新DNA」扭轉缺工困局【專訪】
    Central News Agency (CNA)28 Jun 2026News article

Source summaries

[1]Central News Agency (CNA)28 Jun 2026

CNA interviewed a company executive about AI on the Guan De Wanhua site. The executive described a spray-painting robot run by one operator instead of several workers, cost estimates cut from 3-4 months to 1-2 weeks, and renderings brought in-house at about 1 week. The article also notes that many subcontractors remain hesitant and that all results come from the company.

Full details are in the original sources.

Added 2 Nov 2026 · Updated 2 Nov 2026
Open case page ↗