The construction industry, traditionally known for its reliance on manual processes, is experiencing a technological transformation as artificial intelligence (AI) becomes increasingly integrated into various sectors. During the 2024 Sun Island Annual Conference for Entrepreneurs, Glodon, a leading digital building platform service provider, announced the release of the latest AI models tailored specifically for multiple scenarios in the construction industry.

Li Jiang, Assistant President and General Manager of the Industry AI Department at Glodon, made the announcement at the AI-Driven High-Quality Development Forum. The new L2 AI models, build upon Glodon's previous releases, marks a significant step forward in Glodon’s ongoing efforts to apply AI to real-world construction scenarios, following the launch of its AI large language model for the construction industry, and the AI platform in May.

The L2 models are designed to address specific business scenarios within the construction sector, covering seven key areas: architectural design, quantity take-off, project scheduling, materials management, transaction, safety, and education. These models are tailored to address specific pain points within each domain, offering more targeted and effective solutions. This development signifies a major stride towards more specialized and scenario-driven AI applications in the construction industry.

Pioneering AI Adoption in Construction

Li explained that Glodon's AI model hierarchy consists of three levels: L0, L1, and L2. L0 models refer to general large models trained on general data primarily from the internet; L1 models refer to industry-specific large models that are typically pre-trained or fine-tuned using industry-related data. Mature industry-specific large models are equivalent to AI becoming industry experts, such as Glodon’s AecGPT model launched in May.?

Building upon the general knowledge of L0 models and the professional knowledge of L1 models, L2 models are typically trained for specific problems, and possess knowledge, experience, and workflows in professional scenarios. The core capabilities of these scenario-based models include the ability to automate repetitive tasks, data-driven analysis and decision-making, and understanding and generating professional content.?

Due to its superior graphics engine, more comprehensive knowledge graph, a wider range of software, and broader application scenarios, Glodon has accumulated the necessary data and domain knowledge for building models for specific industry scenarios over an extended period.

Generally, L2 models are more targeted, have better application effects, and are more practical. The L2 models iteratively released by Glodon this time meet the AI requirements of different stages and applications, and truly bring value to construction users.

Li pointed out that the seven L2 models introduced this time all originate from the real-world pain points and difficulties of various business scenarios. They primarily use AI to replace manual labor in completing high-volume, complex tasks in various positions. These large models have very practical application significance for specific groups or positions in specific scenarios.

AI Design Large Model: Liberating Designers' Creativity

Among the many scenario-based large models, the most eye-catching is the scheme design large model. By providing the model with design-related design texts, design drawings, and models, it supports multiple application scenarios such as design specification retrieval, intelligent layout, and design generation.

Glodon, relying on the AI design large model, has developed a new generation of AI-driven scheme design product, CONCETTO (Glodon NovaDesign Scheme Design), which can promote efficient design and innovation, and improve design rationality and value through intelligent generative design and digital decision support. It assists architects in quickly and accurately completing the most time-consuming and error-prone evaluation, listing, and mapping tasks in daily design work, thereby significantly improving design efficiency and optimizing project timelines and costs. This product is currently open for trial use by architects.

AI Quantity Take-off Large Model: Automatic Intelligent Calculation Saves Time and Improves Efficiency

Quantity take-off is very time-consuming in the construction cost engineering field. By deconstructing the quantity take-off workflow, Glodon found that some parts of each workflow can be automated and replaced by AI. For this purpose, Glodon has developed a quantity take-off large model that trains engineering drawings, quantity tables, and engineering documents related to quantity take-off into the large model. Through the large model's layout analysis and multimodal reasoning capabilities, it supports applications in different drawings, tables, and text extraction, thereby significantly reducing the time for quantity take-off.

Glodon's new generation of infrastructure quantity take-off products is based on the quantity take-off large model. In the traditional quantity take-off mode, a highway project would undergo centralized engineering quantity review before officially entering the site, which typically takes about 3 months for a 1 billion yuan project, consuming both time and effort. Using AI quantity take-off, the time for a 1 billion yuan project can be reduced from 3 months to 2 weeks, significantly improving work efficiency.

AI Zebra Plan Software Large Model: Automatic Compilation of Progress Plans

The on-time completion rate in the construction industry is only 23%. For example, for a 100,000-square-meter project, a one-day delay in progress would result in an average cost loss of 42,600 yuan. Progress is the biggest cost. Glodon utilizes the AI progress large model to create a new generation of the Zebra Plan Platform. By training progress management-related text and BIM models into the model, it can extract information from tens of thousands of words of long documents and millions of elements, such as accurately recalling key text information from documents that combine a large amount of text, tables, and images, and output the required engineering information according to formatted requirements.

The AI planning assistant adopts the copilot mode, greatly simplifying the plan compilation process. In the process of bidding plan compilation and construction plan formulation, the plan compilation time is shortened to 30 minutes, improving the project completion rate and achieving zero delays at key project nodes.

AI Material Management Large Model: Empowering Refine Management and Control of Material Costs

Efficient material management is a key factor in controlling project costs, especially since frontline tasks like receiving and dispatching materials involve a large workload. To address the urgent need for increased efficiency on-site, Glodon has developed a material management large model. This model processes and trains data from images, delivery slips, license plates, materials, and other relevant information at the job site. By leveraging the model's capabilities in layout analysis, semantic understanding, and object detection and recognition, it enables AI-powered applications such as recognizing delivery slips, identifying materials and vehicles, and counting materials, with an accuracy exceeding 95%.

Based on this material management large model, Glodon has launched a new generation of intelligent material management systems to help improve the efficiency of material entry and exit weighing, reduce labor costs, and improve management through video visualization and data connectivity, providing tools for personnel assessment. This intelligent control helps projects reduce material losses by 1%.

AI Transaction Large Model: Facilitating Intelligent Automation of the Entire Engineering Construction Bidding, Tendering and Evaluation Process

The transparency and standardization of bidding and tendering have always been the common needs of regulatory bodies, trading departments, bidders and tenderers. Glodon builds an AI transaction large model platform to assist in intelligent transactions. By organizing and training relevant bidding documents, qualification bids, economic bids, technical bids, and construction industry templates into the large model, and applying the large model's semantic understanding, rule reasoning, and content generation capabilities, it realizes applications such as bidding rule extraction, evaluation point extraction, format checking, similarity detection, intelligent evaluation, and intelligent generation of bidding and tendering documents.

Glodon's intelligent transaction platform, built on the AI transaction large model, introduces systems such as intelligent bid evaluation assistance. This system uses AI to analyze and extract the requirements from bidding documents, derive review rules, and intelligently analyze bid documents to assess responsiveness. It also supports intelligent technical bid evaluation, allowing AI to help experts quickly locate key information, speeding up the review process and improving efficiency through human-machine collaboration. This AI-assisted bid evaluation not only enhances the efficiency of bidding activities but also strengthens the standardization and fairness of the evaluation process, further regulating market order.

AI Safety Management Large Model: Achieving a Seamless PDCA Cycle for Safety Management

Construction site safety management is crucial, affecting not only workers' lives and health but also project timelines, quality, and overall company profitability. The application of AI technology in construction site safety management not only improves supervision efficiency and reduces accidents but also helps lower labor costs and improve construction efficiency.?

Glodon has developed an AI safety management large model, training it on a vast dataset comprising risk logs, accident reports, construction standards, and on-site videos and images. Leveraging advanced video analysis, image recognition, and natural language understanding, this model can proactively identify potential hazards, detect safety issues, assess construction risks, and provide access to safety regulations.

AI Education Large Model: Nurturing Project Managers

Glodon has created an AI education large model by training it on accumulated industry experience, regulations, standard blueprints, and professional exam materials. This model has significantly outperformed general models in China’s Level 1 Constructor Certification exams, even surpassing the average human performance. China’s Level 1 Constructor Certification exams, jointly organized by China’s Ministry of Housing and Urban-Rural Development and the Ministry of Human Resources and Social Security, is a professional qualification exam for construction project managers and a prerequisite for becoming a project manager on large-scale projects. Historically, the pass rate of this exam has been between 7% and 10%, and the total number of certified professionals nationwide is only around 30,000.

Glodon has developed an AI project assistant named Tiantian Xiangshang (which translates to Striving for Daily Improvement) based on the AI education large model. This assistant provides project managers with typical case studies, experience sharing, and knowledge aggregation, helping them deepen their industry knowledge. Utilizing AI technology and expert discussions, it aims to assist project managers in quickly obtaining high-quality information and resolving daily challenges. This app is already being used in various key projects and industry competitions across different regions.

Beyond the seven L2 scenario-based large models, Glodon has also developed an AI platform for the construction industry. This platform is a one-stop AI native application development platform, comprising three layers: a large language model for the construction industry layer, a tool platform layer, and a product application layer. It is open to industry partners, enabling them to develop AI applications and promote the AI-driven upgrading of construction industry software and applications, ultimately creating a super brain for the construction industry.

Looking ahead, Glodon aims to continue driving the integration of AI into various construction industry scenarios, promoting the industry’s development towards greater industrialization, digitalization, and sustainability, while helping construction companies improve quality, efficiency, and cost-effectiveness.