How C3 AI helped cut model validation time, scale operations, and unlock $65M in value

By Will Fleischer, AI Solution Manager, C3 AI and Youssef Aitousarrah, Lead Data Scientist, C3 AI

Engineering and construction projects often face significant operations challenges, with 20% running late and 80% exceeding budgets. Generative AI offers a solution that transforms how teams plan, collaborate, and execute by streamlining workflows and improving efficiency.

The building design phase is often where we find these roadblocks in construction projects. One of the most resource-intensive steps is the creation and validation of 3D building design models. This process requires modelers to translate complex engineering designs into 3D models and ensure their accuracy before construction begins. It involves synthesizing data from unstructured documents (e.g., annotated contracts) and structured databases, which must be cross-referenced to meet project specifications.

These labor-intensive steps can take weeks and delay project timelines, but generative AI dramatically reduces this burden by automating critical tasks, enabling teams to focus on higher-value activities.

Streamlining 3D Modeling and Simulation with C3 Generative AI

Generative AI streamlines the detailing process by automatically extracting key details from unstructured and structured documents, enabling modelers to create and validate 3D building design models faster and with greater accuracy. This eliminates tedious manual steps, significantly reducing project timelines. Advances in generative AI are revolutionizing how companies interact with their data. Intelligent agents and large language models (LLMs) can now autonomously retrieve relevant information, generate actionable insights, and orchestrate workflows to deliver measurable results.

C3 Generative AI harnesses these innovations with an LLM-agnostic architecture, selecting the best model for tasks such as analyzing images, identifying patterns, and translating natural language queries into precise, multi-step solutions. Unlike traditional AI solutions, C3 Generative AI ensures enterprise-grade security, compliance, and governance, allowing teams to adopt this cutting-edge technology with confidence. The result is a secure, scalable solution that transforms how building design teams operate, unlocking new levels of efficiency and precision.

A leading building design firm faced significant challenges in validating 3D building design models, a critical step in their workflow. On average, the creation and validation process for each model required approximately 200 hours per project. Modelers were burdened with manually searching through numerous documents to locate relevant information — a painstaking, repetitive task that created bottlenecks and slowed project timelines.

Recognizing the need for a more efficient approach, the company sought a solution to reduce the time and effort required for model validation while ensuring accuracy. C3 Generative AI for Modeling and Simulation transformed the validation process by automating information retrieval and enhancing discrepancy detection between source documents and 3D models. The application streamlined the workflow by flagging inconsistencies, enabling modelers to spend their time resolving issues instead of identifying them. This not only accelerated the validation process but also significantly improved accuracy and operational efficiency.

In practice, the firm reduced validation time by four hours per building across approximately 7,000 projects annually. By eliminating this bottleneck and increasing operational capacity, C3 Generative AI is projected to deliver $65M in annual economic value, empowering the company to scale operations without compromising precision.

The C3 Agentic AI Platform transformed the team’s data management by integrating seven disparate data sources into a unified system. This integration enabled users to access and interact with the most relevant information from both structured and unstructured data through an intuitive interface. By employing an object-oriented programming model, the platform simplified the development of new features, addressing complex challenges like preserving hierarchical relationships within projects while seamlessly scaling to support thousands of projects annually.

Leveraging the platform’s distributed computing capabilities, the solution efficiently handles concurrent building design analyses, ensuring responsive performance at scale. Automated data integration workflows and seamless file system interactions further enhanced efficiency and reliability.

Additionally, the platform’s LLM-agnostic architecture empowered users to experiment with multiple large language models within the same application. This flexibility allowed the team to quickly test, combine, and customize AI models to address the specific building design use case, driving rapid innovation and operational scalability.

Inside the Solution: Data, Parsing, and Scalability

Data Integration and Awareness

Data Integration: Consolidated years of data from seven distinct sources, including both structured and unstructured formats spanning over 50 tables and 2,000 documents across two facilities.Document Complexity: Processed complex PDFs detailing various building components, including multi-page tables, annotated sections, horizontally split columns, and free-text fields.Contract Amendment Coverage: Included documents with updates or modifications to original contracts, ensuring the retrieval of accurate and most current information.Streamlined Data Retrieval: All structured data sources are seamlessly ingested into the C3 AI data model. Relevant insights, such as 3D building design model data, are extracted to address nearly 100 predefined questions efficiently.

Parsing and Vision Language Models (VLMs)

Contract Ingestion and Parsing: C3 Generative AI for Modeling and Simulation processed contracts and amendments in PDF format using a custom parsing approach to extract critical insights. This included transforming contract data into images for enhanced analysis.Advanced Image Analysis with VLMs: Leveraged vision language models to reason over the extracted images and provide precise answers to in-scope questions based on the content.Version Control and Amendment Prioritization: Ensured that the most recent contract amendments took precedence, resolving any conflicting information from earlier versions to maintain data accuracy and consistency.

Scalability

Distributed Architecture: Designed and implemented distributed computing logic to handle computations efficiently, enabling the application to scale seamlessly across the customer’s organization.

Revolutionizing Engineering with Scalable AI-Driven Precision and Efficiency

At scale, the application is projected to generate $65M in annual economic value by increasing operational capacity and reducing quality issues. By streamlining information retrieval from contracts, the solution enables teams to efficiently create, compare, and validate 3D building design models. This enhanced process saves an average of four hours per building, optimizing workflows and freeing up resources to tackle additional projects, driving greater productivity and scalability.

C3 Generative AI for Modeling and Simulation has proven its ability to dramatically reduce engineering design and validation time while enhancing overall operational efficiency. By streamlining workflows and automating time-intensive tasks like data retrieval and validation, the application empowers engineering teams to focus on high-value activities such as innovation and optimization.

This transformative capability enables organizations to meet tight project timelines, reduce errors, and scale their operations effectively. As industries increasingly adopt AI-powered solutions, C3 Generative AI sets a new standard for driving productivity, precision, and measurable economic value across engineering and design functions.

About the Authors 

Will Fleischer is an AI Solution Manager on the Generative AI team at C3 AI, where he leads cross functional teams of Data Scientists and Engineers working to develop Generative AI pilots for various industries. He holds a bachelor’s degree from University of Pittsburgh.

Youssef Aitousarrah is a Lead Data Scientist on the Generative AI Data Science team at C3 AI, where he leads and develops advanced applications to solve complex industrial challenges in multiple domains like energy and healthcare. His current work focuses on leveraging agentic frameworks to build performant, robust, and scalable generative AI solutions to increase productivity across industries and governmental agencies. He holds a MSc in Energy Resources Engineering from Stanford University and a Diplome d’Ingenieur from the prestigious Ecole Polytechnique in Paris.