"ChatGPT Lies, But We Are 100% Accurate"

A single design error can cost hundreds of billions of won. This is a chronic problem in the Middle East construction industry. In mega-projects worth trillions of won, such as Saudi Arabia's NEOM City, even minor design mistakes can escalate into losses of hundreds of billions of won. This is a challenge that even McKinsey, with its 10,000-strong AI team, failed to solve—and now Korean AI startup Crimson Amber is taking it on.

CEO Woo Je-geun states, "ChatGPT lies when it comes to design blueprints. We designed our system so that it cannot lie." The secret lies in ontology-based BIM (Building Information Modeling) technology—a system that weaves together construction laws, design standards, and BIM models into a knowledge graph to detect errors.

Design Errors Cost Hundreds of Billions

According to a U.S. Construction Association survey, design errors account for 56.5% of construction cost overruns and 40% of project delays. "Middle East projects are so large in scale that small mistakes get amplified. A single structural calculation error can affect hundreds of columns and bring the entire construction process to a halt."

The problem is that humans must manually reconcile consistency across countless regulations, specifications, and drawings. Fatigue accumulates during this process, which takes months, leading to mistakes. "This gap is the market we're penetrating." The goal is to overcome design errors with AI and achieve a 5-10% reduction in construction costs.

Knowledge Graphs Create 100% Accurate Answers

Crimson Amber's solution under development, 'Artisan Construction,' combines a chat interface with a 3D BIM viewer. "We uploaded a design blueprint to GPT-4o and asked, 'How many rooms are there?' It got it wrong. LLMs cannot understand the structural relationships in design documents."

The solution is ontology and knowledge graphs. The system interprets BIM models to extract objects like walls, columns, and slabs, along with their relationships. This information is stored in a graph structure, overlaid with construction laws and design standards modeled as ontologies.

For example, if there's a regulation stating "evacuation stairs must be wider than a certain width," the AI traces the graph to read the actual width and compares it against the standard to make a determination. "Our AI doesn't guess. It only answers based on facts defined in the knowledge graph." Academic credibility has been secured through joint research with the Korea Institute of Civil Engineering and Building Technology.

Technology First Validated by the Korean Government

In 2022, the Ministry of Land, Infrastructure and Transport began mandating BIM for road construction projects exceeding 100 billion won. By 2030, this will expand to all public construction projects. The 'Construction Standards Digitalization' project is also underway to convert 1,079 construction standards into ontology format.

"The technology the government aims to complete by 2026, we have already commercialized. What's happening in Korea right now is the perfect testbed for export to the Middle East." The strategy is to first establish the system as a standard tool in Korean projects. The explanation that it is "AI infrastructure validated by the Korean government and major construction companies" carries persuasive power with Middle East clients.
 

NEOM City is the Goal... "5-10% Reduction in Construction Costs"

When CEO Woo discusses the Middle East, his keyword is ROI. "By simply reducing errors in the design phase, we can save 5-10% of total construction costs. For a 1 trillion won project, that's 50 to 100 billion won in savings."

The NEOM City scenario is specific. First, the ontology BIM AI validated in Korea will be introduced as a pilot in specific NEOM City projects. International codes and local standards will be modeled as ontologies for rapid localization. Second is joint expansion with Korean construction companies. Leveraging the reputation built by Hyundai E&C, Samsung C&T, and GS E&C, the company will propose packages that automate design review, permitting, and client reporting. Third is platformization—expanding project-by-project accumulated data into a shared knowledge infrastructure.

"Exporting Korea's 2030 Digital Construction Experience to the Middle East"

"Our goal is to experience Korea's construction digitalization through 2030, then export that know-how to the Middle East. Our vision is to make the 'hallucination-free construction AI platform' validated in Korea the common infrastructure for global infrastructure projects, including NEOM City."

CEO Woo, who majored in physics, earned an MBA, and accumulated 12 years of experience in finance and pharmaceutical companies, entered the construction AI market drawing on his diverse industry background. Crimson Amber plans to launch Artisan Construction V1.0 as a paid service in March 2026. "We're creating a new category of AI specialized in construction blueprint interpretation, like AlphaFold." A company doing what ChatGPT cannot—Crimson Amber's challenge to the Middle East construction market with precision is gaining attention.